Electromagnetics, Volume 2 1949373916, 9781949373912

Free online at: https://doi.org/10.21061/electromagnetics-vol-2 Electromagnetics, volume 2 by Steven W. Ellingson is a 2

305 37 5MB

English Pages 228 [230] Year 2019

Report DMCA / Copyright

DOWNLOAD PDF FILE

Recommend Papers

Electromagnetics, Volume 2
 1949373916, 9781949373912

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

ELECTROMAGNETICS STEVEN W. ELLINGSON

VOLUME 2

ELECTROMAGNETICS

VOLUME 2

Publication of this book was made possible in part by the Virginia Tech University Libraries’ Open Education Initiative Faculty Grant program: http://guides.lib.vt.edu/oer/grants

The Open Electromagnetics Project, https://www.faculty.ece.vt.edu/swe/oem

Books in this series Electromagnetics, Volume 1, https://doi.org/10.21061/electromagnetics-vol-1 Electromagnetics, Volume 2, https://doi.org/10.21061/electromagnetics-vol-2

ELECTROMAGNETICS STEVEN W. ELLINGSON VOLUME 2

Copyright © 2020 Steven W. Ellingson This textbook is licensed with a Creative Commons Attribution Share-Alike 4.0 license: https://creativecommons.org/licenses/by-sa/4.0. You are free to copy, share, adapt, remix, transform, and build upon the material for any purpose, even commercially, as long as you follow the terms of the license: https://creativecommons.org/licenses/by-sa/4.0/legalcode. This work is published by Virginia Tech Publishing, a division of the University Libraries at Virginia Tech, 560 Drillfield Drive, Blacksburg, VA 24061, USA ([email protected]). Suggested citation: Ellingson, Steven W. (2020) Electromagnetics, Vol. 2. Blacksburg, VA: Virginia Tech Publishing. https://doi.org/10.21061/electromagnetics-vol-2. Licensed with CC BY-SA 4.0. https://creativecommons.org/licenses/bysa/4.0. Peer Review: This book has undergone single-blind peer review by a minimum of three external subject matter experts. Accessibility Statement: Virginia Tech Publishing is committed to making its publications accessible in accordance with the Americans with Disabilities Act of 1990. The screen reader–friendly PDF version of this book is tagged structurally and includes alternative text which allows for machine-readability. The LaTeX source files also include alternative text for all images and figures. Publication Cataloging Information Ellingson, Steven W., author Electromagnetics (Volume 2) / Steven W. Ellingson Pages cm ISBN 978-1-949373-91-2 (print) ISBN 978-1-949373-92-9 (ebook) DOI: https://doi.org/10.21061/electromagnetics-vol-2 1. Electromagnetism. 2. Electromagnetic theory. I. Title QC760.E445 2020 621.3 The print version of this book is printed in the United States of America. Cover Design: Robert Browder Cover Image: © Michelle Yost. Total Internal Reflection (https://flic.kr/p/dWAhx5) is licensed with a Creative Commons Attribution-ShareAlike 2.0 license: https://creativecommons.org/licenses/by-sa/2.0/ (cropped by Robert Browder)

iv

Features of This Open Textbook Additional Resources

The following resources are freely available at http://hdl.handle.net/10919/93253 Print edition ordering details Downloadable PDF of the book Links to collaborator portal and listserv LaTeX source files Links to other books in the series Slides of figures used in the book Errata Problem sets and solution manual

Review / Adopt /Adapt / Build upon

If you are an instructor reviewing, adopting, or adapting this textbook, please help us understand your use by completing this form: http://bit-ly/vtpublishing-update. You are free to copy, share, adapt, remix, transform, and build upon the material for any purpose, even commercially, as long as you follow the terms of the license: https://creativecommons.org/licenses/by-sa/4.0/legalcode. You must: Attribute — You must give appropriate credit, provide a link to the license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use. Suggested citation: Adapted by _[your name]_ from (c) Steven W. Ellingson, Electromagnetics, Vol 2, https://doi .org/10.21061/electromagnetics-vol-2, CC BY SA 4.0, https://creativecommons.org/licenses/by-sa/4.0. ShareAlike — If you remix, transform, or build upon the material, you must distribute your contributions under the same license as the original. You may not: Add any additional restrictions — You may not apply legal terms or technological measures that legally restrict others from doing anything the license permits. If adapting or building upon, you are encouraged to: Incorporate only your own work or works with a CC BY or CC BY-SA license. Attribute all added content. Have your work peer reviewed. Include a transformation statement that describes changes, additions, accessibility features, and any subsequent peer review. If incorporating text or figures under an informed fair use analysis, mark them as such and cite them. Share your contributions in the collaborator portal or on the listserv. Contact the author to explore contributing your additions to the project. Adaptation resources LaTeX source files are available. Adapt on your own at https://libretexts.org. Guide: Modifying an Open Textbook https://press.rebus.community/otnmodify.

Suggestions for creating and adapting Create and share learning tools and study aids. Translate. Modify the sequence or structure. Add or modify problems and examples. Transform or build upon in other formats.

Submit suggestions and comments Submit suggestions (anonymous): http://bit.ly/electromagnetics-suggestion Email: [email protected] Annotate using Hypothes.is http://web.hypothes.is For more information see the User Feedback Guide: http://bit.ly/userfeedbackguide

v

Contents Preface 1

ix

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

1 1 3 4 5

Magnetostatics Redux 2.1 Lorentz Force . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Magnetic Force on a Current-Carrying Wire . . . . . . . . 2.3 Torque Induced by a Magnetic Field . . . . . . . . . . . . 2.4 The Biot-Savart Law . . . . . . . . . . . . . . . . . . . . 2.5 Force, Energy, and Potential Difference in a Magnetic Field

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

11 11 12 15 18 20

Wave Propagation in General Media 3.1 Poynting’s Theorem . . . . . . . . 3.2 Poynting Vector . . . . . . . . . . 3.3 Wave Equations for Lossy Regions 3.4 Complex Permittivity . . . . . . . 3.5 Loss Tangent . . . . . . . . . . . 3.6 Plane Waves in Lossy Regions . . 3.7 Wave Power in a Lossy Medium . 3.8 Decibel Scale for Power Ratio . . 3.9 Attenuation Rate . . . . . . . . . 3.10 Poor Conductors . . . . . . . . . 3.11 Good Conductors . . . . . . . . . 3.12 Skin Depth . . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

25 25 28 30 33 34 36 37 39 40 41 43 46

4

Current Flow in Imperfect Conductors 4.1 AC Current Flow in a Good Conductor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Impedance of a Wire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Surface Impedance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

48 48 50 54

5

Wave Reflection and Transmission 5.1 Plane Waves at Normal Incidence on a Planar Boundary . . . . . 5.2 Plane Waves at Normal Incidence on a Material Slab . . . . . . 5.3 Total Transmission Through a Slab . . . . . . . . . . . . . . . . 5.4 Propagation of a Uniform Plane Wave in an Arbitrary Direction .

56 56 60 64 67

2

3

Preliminary Concepts 1.1 Units . . . . . . . . . . . . . . . . . . . . 1.2 Notation . . . . . . . . . . . . . . . . . . 1.3 Coordinate Systems . . . . . . . . . . . . 1.4 Electromagnetic Field Theory: A Review

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . .

. . . . . . . . . . . .

vi

. . . .

. . . . . . . . . . . .

. . . .

. . . . . . . . . . . .

. . . .

. . . . . . . . . . . .

. . . .

. . . . . . . . . . . .

. . . .

. . . . . . . . . . . .

. . . .

. . . . . . . . . . . .

. . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

CONTENTS 5.5 5.6 5.7 5.8 5.9 5.10 5.11 5.12

vii

Decomposition of a Wave into TE and TM Components . . . . . . Plane Waves at Oblique Incidence on a Planar Boundary: TE Case Plane Waves at Oblique Incidence on a Planar Boundary: TM Case Angles of Reflection and Refraction . . . . . . . . . . . . . . . . TE Reflection in Non-magnetic Media . . . . . . . . . . . . . . . TM Reflection in Non-magnetic Media . . . . . . . . . . . . . . . Total Internal Reflection . . . . . . . . . . . . . . . . . . . . . . Evanescent Waves . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

70 72 76 80 83 85 88 90

6 Waveguides 6.1 Phase and Group Velocity . . . . . . . . . . . . . . . 6.2 Parallel Plate Waveguide: Introduction . . . . . . . . 6.3 Parallel Plate Waveguide: TE Case, Electric Field . . 6.4 Parallel Plate Waveguide: TE Case, Magnetic Field . 6.5 Parallel Plate Waveguide: TM Case, Electric Field . . 6.6 Parallel Plate Waveguide: The TM0 Mode . . . . . . 6.7 General Relationships for Unidirectional Waves . . . 6.8 Rectangular Waveguide: TM Modes . . . . . . . . . 6.9 Rectangular Waveguide: TE Modes . . . . . . . . . 6.10 Rectangular Waveguide: Propagation Characteristics

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

95 . 95 . 97 . 99 . 102 . 104 . 107 . 108 . 110 . 113 . 117

7 Transmission Lines Redux 7.1 Parallel Wire Transmission Line . . . . . . . 7.2 Microstrip Line Redux . . . . . . . . . . . . 7.3 Attenuation in Coaxial Cable . . . . . . . . . 7.4 Power Handling Capability of Coaxial Cable . 7.5 Why 50 Ohms? . . . . . . . . . . . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

121 . 121 . 123 . 129 . 133 . 135

. . . . .

. . . . .

. . . . .

. . . . .

8 Optical Fiber 138 8.1 Optical Fiber: Method of Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 8.2 Acceptance Angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 8.3 Dispersion in Optical Fiber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141 9 Radiation 9.1 Radiation from a Current Moment . . . . . . . . . . . . . . 9.2 Magnetic Vector Potential . . . . . . . . . . . . . . . . . . . 9.3 Solution of the Wave Equation for Magnetic Vector Potential 9.4 Radiation from a Hertzian Dipole . . . . . . . . . . . . . . . 9.5 Radiation from an Electrically-Short Dipole . . . . . . . . . 9.6 Far-Field Radiation from a Thin Straight Filament of Current 9.7 Far-Field Radiation from a Half-Wave Dipole . . . . . . . . 9.8 Radiation from Surface and Volume Distributions of Current

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

145 . 145 . 147 . 150 . 152 . 155 . 159 . 161 . 162

10 Antennas 10.1 How Antennas Radiate . . . . . . . . . . . . . . . . . . . . . . 10.2 Power Radiated by an Electrically-Short Dipole . . . . . . . . . 10.3 Power Dissipated by an Electrically-Short Dipole . . . . . . . . 10.4 Reactance of the Electrically-Short Dipole . . . . . . . . . . . . 10.5 Equivalent Circuit Model for Transmission; Radiation Efficiency 10.6 Impedance of the Electrically-Short Dipole . . . . . . . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

166 . 166 . 168 . 169 . 171 . 173 . 175

. . . . . . . .

CONTENTS

viii 10.7 10.8 10.9 10.10 10.11 10.12 10.13 10.14

Directivity and Gain . . . . . . . . . . . . . . . Radiation Pattern . . . . . . . . . . . . . . . . Equivalent Circuit Model for Reception . . . . Reciprocity . . . . . . . . . . . . . . . . . . . Potential Induced in a Dipole . . . . . . . . . . Equivalent Circuit Model for Reception, Redux Effective Aperture . . . . . . . . . . . . . . . . Friis Transmission Equation . . . . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. 177 . 179 . 183 . 186 . 190 . 194 . 196 . 201

A Constitutive Parameters of Some Common Materials 205 A.1 Permittivity of Some Common Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 A.2 Permeability of Some Common Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206 A.3 Conductivity of Some Common Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207 B Mathematical Formulas 209 B.1 Trigonometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209 B.2 Vector Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209 B.3 Vector Identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211 C Physical Constants

212

Index

213

Preface About This Book

• Chapter 5 (“Wave Reflection and Transmission”) addresses scattering of plane waves from planar interfaces.

[m0213]

Goals for this book. This book is intended to serve as a primary textbook for the second semester of a two-semester course in undergraduate engineering electromagnetics. The presumed textbook for the first semester is Electromagnetics Vol. 1,1 which addresses the following topics: electric and magnetic fields; electromagnetic properties of materials; electromagnetic waves; and devices that operate according to associated electromagnetic principles including resistors, capacitors, inductors, transformers, generators, and transmission lines. The book you are now reading – Electromagnetics Vol. 2 – addresses the following topics: • Chapter 1 (“Preliminary Concepts”) provides a brief summary of conventions for units, notation, and coordinate systems, and a synopsis of electromagnetic field theory from Vol. 1. • Chapter 2 (“Magnetostatics Redux”) extends the coverage of magnetostatics in Vol. 1 to include magnetic forces, rudimentary motors, and the Biot-Savart law. • Chapter 3 (“Wave Propagation in General Media”) addresses Poynting’s theorem, theory of wave propagation in lossy media, and properties of imperfect conductors. • Chapter 4 (“Current Flow in Imperfect Conductors”) addresses the frequency-dependent distribution of current in wire conductors and subsequently the AC impedance of wires. 1 S.W.

Ellingson, Electromagnetics Vol. 1, VT Publishing, 2018. CC BY-SA 4.0. ISBN 9780997920185. https://doi.org/10.21061/electromagnetics-vol-1

• Chapter 6 (“Waveguides”) provides an introduction to waveguide theory via the parallel plate and rectangular waveguides. • Chapter 7 (“Transmission Lines Redux”) extends the coverage of transmission lines in Vol. 1 to include parallel wire lines, the theory of microstrip lines, attenuation, and power-handling capabilities. The inevitable but hard-to-answer question “What’s so special about 50 Ω?” is addressed at the end of this chapter. • Chapter 8 (“Optical Fiber”) provides an introduction to multimode fiber optics, including the concepts of acceptance angle and modal dispersion. • Chapter 9 (“Radiation”) provides a derivation of the electromagnetic fields radiated by a current distribution, emphasizing the analysis of line distributions using the Hertzian dipole as a differential element. • Chapter 10 (“Antennas”) provides an introduction to antennas, emphasizing equivalent circuit models for transmission and reception and characterization in terms of directivity and pattern. This chapter concludes with the Friis transmission equation. Appendices covering material properties, mathematical formulas, and physical constants are repeated from Vol. 1, with a few additional items. Target audience. This book is intended for electrical engineering students in the third year of a bachelor of science degree program. It is assumed that students have successfully completed one semester of

c 2020 S.W. Ellingson CC BY SA 4.0. https://doi.org/10.21061/electromagnetics-vol-2 Electromagnetics Vol. 2.

PREFACE

x engineering electromagnetics, nominally using Vol. 1. However, the particular topics and sequence of topics in Vol. 1 are not an essential prerequisite, and in any event this book may be useful as a supplementary reference when a different textbook is used. It is assumed that readers are familiar with the fundamentals of electric circuits and linear systems, which are normally taught in the second year of the degree program. It is also assumed that readers are proficient in basic engineering mathematics, including complex numbers, trigonometry, vectors, partial differential equations, and multivariate calculus. Notation, examples, and highlights. Section 1.2 summarizes the mathematical notation used in this book. Examples are set apart from the main text as follows: Example 0.1. This is an example. “Highlight boxes” are used to identify key ideas as follows: This is a key idea. What are those little numbers in square brackets? This book is a product of the Open Electromagnetics Project. This project provides a large number of sections (“modules”) which are assembled (“remixed”) to create new and different versions of the book. The text “[m0213]” that you see at the beginning of this section uniquely identifies the module within the larger set of modules provided by the project. This identification is provided because different remixes of this book may exist, each consisting of a different subset and arrangement of these modules. Prospective authors can use this identification as an aid in creating their own remixes. Why do some sections of this book seem to repeat material presented in previous sections? In some remixes of this book, authors might choose to eliminate or reorder modules. For this reason, the modules are written to “stand alone” as much as possible. As a result, there may be some redundancy between sections that would not be present in a traditional (non-remixable) textbook. While this may seem awkward to some at first, there are clear

benefits: In particular, it never hurts to review relevant past material before tackling a new concept. And, since the electronic version of this book is being offered at no cost, there is not much gained by eliminating this useful redundancy. Why cite Wikipedia pages as additional reading? Many modules cite Wikipedia entries as sources of additional information. Wikipedia represents both the best and worst that the Internet has to offer. Most educators would agree that citing Wikipedia pages as primary sources is a bad idea, since quality is variable and content is subject to change over time. On the other hand, many Wikipedia pages are excellent, and serve as useful sources of relevant information that is not strictly within the scope of the curriculum. Furthermore, students benefit from seeing the same material presented differently, in a broader context, and with the additional references available as links from Wikipedia pages. We trust instructors and students to realize the potential pitfalls of this type of resource and to be alert for problems. Acknowledgments. Here’s a list of talented and helpful people who contributed to this book: The staff of Virginia Tech Publishing, University Libraries, Virginia Tech: Acquisitions/Developmental Editor & Project Manager: Anita Walz Advisors: Peter Potter, Corinne Guimont Cover, Print Production: Robert Browder Other Virginia Tech contributors: Accessibility: Christa Miller, Corinne Guimont, Sarah Mease Assessment: Anita Walz Virginia Tech Students: Alt text writer: Michel Comer Figure designers: Kruthika Kikkeri, Sam Lally, Chenhao Wang Copyediting: Longleaf Press External reviewers: Randy Haupt, Colorado School of Mines Karl Warnick, Brigham Young University Anonymous faculty member, research university

xi

Also, thanks are due to the students of the Fall 2019 section of ECE3106 at Virginia Tech who used the beta version of this book and provided useful feedback. Finally, we acknowledge all those who have contributed their art to Wikimedia Commons (https://commons.wikimedia.org/) under open licenses, allowing their work to appear as figures in this book. These contributors are acknowledged in figures and in the “Image Credits” section at the end of each chapter. Thanks to each of you for your selfless effort.

About the Open Electromagnetics Project [m0148]

The Open Electromagnetics Project (https://www.faculty.ece.vt.edu/swe/oem/) was established at Virginia Tech in 2017 with the goal of creating no-cost openly-licensed textbooks for courses in undergraduate engineering electromagnetics. While a number of very fine traditional textbooks are available on this topic, we feel that it has become unreasonable to insist that students pay hundreds of dollars per book when effective alternatives can be provided using modern media at little or no cost to the student. This project is equally motivated by the desire for the freedom to adopt, modify, and improve educational resources. This work is distributed under a Creative Commons BY SA license which allows – and we hope encourages – others to adopt, modify, improve, and expand the scope of our work.

xii

About the Author [m0153]

Steven W. Ellingson ([email protected]) is an Associate Professor at Virginia Tech in Blacksburg, Virginia, in the United States. He received PhD and MS degrees in Electrical Engineering from the Ohio State University and a BS in Electrical and Computer Engineering from Clarkson University. He was employed by the U.S. Army, Booz-Allen & Hamilton, Raytheon, and the Ohio State University ElectroScience Laboratory before joining the faculty of Virginia Tech, where he teaches courses in electromagnetics, radio frequency electronics, wireless communications, and signal processing. His research includes topics in wireless communications, radio science, and radio frequency instrumentation. Professor Ellingson serves as a consultant to industry and government and is the author of Radio Systems Engineering (Cambridge University Press, 2016) and Electromagnetics Vol. 1 (VT Publishing, 2018).

PREFACE

Chapter 1

Preliminary Concepts 1.1 Units [m0072]

The term “unit” refers to the measure used to express a physical quantity. For example, the mean radius of the Earth is about 6,371,000 meters; in this case, the unit is the meter. A number like “6,371,000” becomes a bit cumbersome to write, so it is common to use a prefix to modify the unit. For example, the radius of the Earth is more commonly said to be 6371 kilometers, where one kilometer is understood to mean 1000 meters. It is common practice to use prefixes, such as “kilo-,” that yield values in the range of 0.001 to 10, 000. A list of standard prefixes appears in Table 1.1.

Prefix exa peta tera giga mega kilo milli micro nano pico femto atto

Abbreviation E P T G M k m µ n p f a

Multiply by: 1018 1015 1012 109 106 103 10−3 10−6 10−9 10−12 10−15 10−18

Table 1.1: Prefixes used to modify units.

Unit ampere coulomb farad henry hertz joule meter newton ohm second tesla volt watt weber

Abbreviation A C F H Hz J m N Ω s T V W Wb

Quantifies: electric current electric charge capacitance inductance frequency energy distance force resistance time magnetic flux density electric potential power magnetic flux

Table 1.2: Some units that are commonly used in electromagnetics. Writing out the names of units can also become tedious. For this reason, it is common to use standard abbreviations; e.g., “6731 km” as opposed to “6371 kilometers,” where “k” is the standard abbreviation for the prefix “kilo” and “m” is the standard abbreviation for “meter.” A list of commonly-used base units and their abbreviations are shown in Table 1.2. To avoid ambiguity, it is important to always indicate the units of a quantity; e.g., writing “6371 km” as opposed to “6371.” Failure to do so is a common source of error and misunderstandings. An example is the expression: l = 3t where l is length and t is time. It could be that l is in

c 2020 S.W. Ellingson CC BY SA 4.0. https://doi.org/10.21061/electromagnetics-vol-2 Electromagnetics Vol. 2.

CHAPTER 1. PRELIMINARY CONCEPTS

2 meters and t is in seconds, in which case “3” really means “3 m/s.” However, if it is intended that l is in kilometers and t is in hours, then “3” really means “3 km/h,” and the equation is literally different. To patch this up, one might write “l = 3t m/s”; however, note that this does not resolve the ambiguity we just identified – i.e., we still don’t know the units of the constant “3.” Alternatively, one might write “l = 3t where l is in meters and t is in seconds,” which is unambiguous but becomes quite awkward for more complicated expressions. A better solution is to write instead: l = (3 m/s) t

(mol), and luminosity in candela (cd). SI units for electromagnetic quantities such as coulombs (C) for charge and volts (V) for electric potential are derived from these fundamental units.

or even better:

Additional Reading:

A frequently-overlooked feature of units is their ability to assist in error-checking mathematical expressions. For example, the electric field intensity may be specified in volts per meter (V/m), so an expression for the electric field intensity that yields units of V/m is said to be “dimensionally correct” (but not necessarily correct), whereas an expression that cannot be reduced to units of V/m cannot be correct.

l = at where a = 3 m/s since this separates the issue of units from the perhaps more-important fact that l is proportional to t and the constant of proportionality (a) is known. The meter is the fundamental unit of length in the International System of Units, known by its French acronym “SI” and sometimes informally referred to as the “metric system.” In this work, we will use SI units exclusively. Although SI is probably the most popular for engineering use overall, other systems remain in common use. For example, the English system, where the radius of the Earth might alternatively be said to be about 3959 miles, continues to be used in various applications and to a lesser or greater extent in various regions of the world. An alternative system in common use in physics and material science applications is the CGS (“centimeter-gram-second”) system. The CGS system is similar to SI, but with some significant differences. For example, the base unit of energy in the CGS system is not the “joule” but rather the “erg,” and the values of some physical constants become unitless. Therefore – once again – it is very important to include units whenever values are stated. SI defines seven fundamental units from which all other units can be derived. These fundamental units are distance in meters (m), time in seconds (s), current in amperes (A), mass in kilograms (kg), temperature in kelvin (K), particle count in moles

• “International System of Units” on Wikipedia. • “Centimetre-gram-second system of units” on Wikipedia.

1.2. NOTATION

3 Z

1.2 Notation [m0005]

The list below describes notation used in this book. • Vectors: Boldface is used to indicate a vector; e.g., the electric field intensity vector will typically appear as E. Quantities not in boldface are scalars. When writing by hand, it is common → − to write “E” or “ E ” in lieu of “E.” • Unit vectors: A circumflex is used to indicate a unit vector; i.e., a vector having magnitude equal to one. For example, the unit vector pointing in ˆ . In the +x direction will be indicated as x discussion, the quantity “ˆ x” is typically spoken “x hat.” • Time: The symbol t is used to indicate time. • Position: The symbols (x, y, z), (ρ, φ, z), and (r, θ, φ) indicate positions using the Cartesian, cylindrical, and spherical coordinate systems, respectively. It is sometimes convenient to express position in a manner which is independent of a coordinate system; in this case, we typically use the symbol r. For example, ˆx + y ˆy + z ˆz in the Cartesian coordinate r=x system. • Phasors: A tilde is used to indicate a phasor quantity; e.g., a voltage phasor might be indicated as Ve , and the phasor representation of e E will be indicated as E.

• Curves, surfaces, and volumes: These geometrical entities will usually be indicated in script; e.g., an open surface might be indicated as S and the curve bounding this surface might be indicated as C. Similarly, the volume enclosed by a closed surface S may be indicated as V. • Integrations over curves, surfaces, and volumes will usually be indicated using a single integral sign with the appropriate subscript. For example: Z · · · dl is an integral over the curve C C

Z

S

· · · ds

is an integral over the surface S

V

· · · dv

is an integral over the volume V.

• Integrations over closed curves and surfaces will be indicated using a circle superimposed on the integral sign. For example: I · · · dl is an integral over the closed curve C C

I

S

··· ds is an integral over the closed surface S

A “closed curve” is one which forms an unbroken loop; e.g., a circle. A “closed surface” is one which encloses a volume with no openings; e.g., a sphere. • The symbol “∼ =” means “approximately equal to.” This symbol is used when equality exists, but is not being expressed with exact numerical precision. For example, the ratio of the circumference of a circle to its diameter is π, where π ∼ = 3.14. • The symbol “≈” also indicates “approximately equal to,” but in this case the two quantities are unequal even if expressed with exact numerical precision. For example, ex = 1 + x + x2 /2 + ... as an infinite series, but ex ≈ 1 + x for x ≪ 1. Using this approximation, e0.1 ≈ 1.1, which is in good agreement with the actual value e0.1 ∼ = 1.1052. • The symbol “∼” indicates “on the order of,” which is a relatively weak statement of equality indicating that the indicated quantity is within a factor of 10 or so of the indicated value. For example, µ ∼ 105 for a class of iron alloys, with exact values being larger or smaller by a factor of 5 or so. • The symbol “,” means “is defined as” or “is equal as the result of a definition.” √ • Complex numbers: j , −1. • See Appendix C for notation used to identify commonly-used physical constants.

CHAPTER 1. PRELIMINARY CONCEPTS

4

1.3 Coordinate Systems z [m0180]

z

The coordinate systems most commonly used in engineering analysis are the Cartesian, cylindrical, and spherical systems. These systems are illustrated in Figures 1.1, 1.2, and 1.3, respectively. Note that the use of variables is not universal; in particular, it is common to encounter the use of r in lieu of ρ for the radial coordinate in the cylindrical system, and the use of R in lieu of r for the radial coordinate in the spherical system.

ϕ ρ y z ϕ

ρ

Additional Reading:

x • “Cylindrical coordinate system” on Wikipedia. • “Spherical coordinate system” on Wikipedia.

c K. Kikkeri CC BY SA 4.0

Figure 1.2: Cylindrical coordinate system.

z

ϕ θ

r θ

r y

z

z y x y

ϕ

x x c K. Kikkeri CC BY SA 4.0

Figure 1.1: Cartesian coordinate system.

c K. Kikkeri CC BY SA 4.0

Figure 1.3: Spherical coordinate system.

1.4. ELECTROMAGNETIC FIELD THEORY: A REVIEW

1.4 Electromagnetic Field Theory: A Review

5

This is referred to as Kirchoff’s voltage law for electrostatics. The inverse of Equation 1.1 is E = −∇V

[m0179]

This book is the second in a series of textbooks on electromagnetics. This section presents a summary of electromagnetic field theory concepts presented in the previous volume.

(1.3)

That is, the electric field intensity points in the direction in which the potential is most rapidly decreasing, and the magnitude is equal to the rate of change in that direction.

The flux interpretation of the electric field is referred to as electric flux density D (SI base units of C/m2 ), Electric charge and current. Charge is the ultimate and quantifies the effect of charge as a flow emanating source of the electric field and has SI base units of from the charge. Gauss’ law for electric fields states coulomb (C). An important source of charge is the that the electric flux through a closed surface is equal electron, whose charge is defined to be negative. However, the term “charge” generally refers to a large to the enclosed charge Qencl ; i.e., I number of charge carriers of various types, and whose D · ds = Qencl (1.4) relative net charge may be either positive or negative. S Distributions of charge may alternatively be expressed in terms of line charge density ρl (C/m), surface charge density ρs (C/m2 ), or volume charge Within a material region, we find density ρv (C/m3 ). Electric current describes the net D = ǫE (1.5) motion of charge. Current is expressed in SI base units of amperes (A) and may alternatively be where ǫ is the permittivity (SI base units of F/m) of quantified in terms of surface current density Js the material. In free space, ǫ is equal to (A/m) or volume current density J (A/m2 ). Electrostatics. Electrostatics is the theory of the electric field subject to the constraint that charge does not accelerate. That is, charges may be motionless (“static”) or move without acceleration (“steady current”). The electric field may be interpreted in terms of energy or flux. The energy interpretation of the electric field is referred to as electric field intensity E (SI base units of N/C or V/m), and is related to the energy associated with charge and forces between charges. One finds that the electric potential (SI base units of V) over a path C is given by Z V = − E · dl (1.1) C

The principle of independence of path means that only the endpoints of C in Equation 1.1, and no other details of C, matter. This leads to the finding that the electrostatic field is conservative; i.e., I E · dl = 0 (1.2) C

ǫ0 , 8.854 × 10−12 F/m

(1.6)

It is often convenient to quantify the permittivity of material in terms of the unitless relative permittivity ǫr , ǫ/ǫ0 . Both E and D are useful as they lead to distinct and independent boundary conditions at the boundary between dissimilar material regions. Let us refer to these regions as Regions 1 and 2, having fields (E1 , D1 ) and (E2 , D2 ), respectively. Given a vector ˆ perpendicular to the boundary and pointing into n Region 1, we find ˆ × [E1 − E2 ] = 0 n

(1.7)

i.e., the tangential component of the electric field is continuous across a boundary, and ˆ · [D1 − D2 ] = ρs n

(1.8)

i.e., any discontinuity in the normal component of the electric field must be supported by a surface charge distribution on the boundary.

CHAPTER 1. PRELIMINARY CONCEPTS

6 Magnetostatics. Magnetostatics is the theory of the magnetic field in response to steady current or the intrinsic magnetization of materials. Intrinsic magnetization is a property of some materials, including permanent magnets and magnetizable materials. Like the electric field, the magnetic field may be quantified in terms of energy or flux. The flux interpretation of the magnetic field is referred to as magnetic flux density B (SI base units of Wb/m2 ), and quantifies the field as a flow associated with, but not emanating from, the source of the field. The magnetic flux Φ (SI base units of Wb) is this flow measured through a specified surface. Gauss’ law for magnetic fields states that I B · ds = 0 (1.9) S

i.e., the magnetic flux through a closed surface is zero. Comparison to Equation 1.4 leads to the conclusion that the source of the magnetic field cannot be localized; i.e., there is no “magnetic charge” analogous to electric charge. Equation 1.9 also leads to the conclusion that magnetic field lines form closed loops. The energy interpretation of the magnetic field is referred to as magnetic field intensity H (SI base units of A/m), and is related to the energy associated with sources of the magnetic field. Ampere’s law for magnetostatics states that I H · dl = Iencl (1.10) C

where Iencl is the current flowing past any open surface bounded by C. Within a homogeneous material region, we find

Both B and H are useful as they lead to distinct and independent boundary conditions at the boundaries between dissimilar material regions. Let us refer to these regions as Regions 1 and 2, having fields (B1 , H1 ) and (B2 , H2 ), respectively. Given a vector ˆ perpendicular to the boundary and pointing into n Region 1, we find ˆ · [B1 − B2 ] = 0 n

(1.13)

i.e., the normal component of the magnetic field is continuous across a boundary, and ˆ × [H1 − H2 ] = Js n

(1.14)

i.e., any discontinuity in the tangential component of the magnetic field must be supported by current on the boundary. Maxwell’s equations. Equations 1.2, 1.4, 1.9, and 1.10 are Maxwell’s equations for static fields in integral form. As indicated in Table 1.3, these equations may alternatively be expressed in differential form. The principal advantage of the differential forms is that they apply at each point in space (as opposed to regions defined by C or S), and subsequently can be combined with the boundary conditions to solve complex problems using standard methods from the theory of differential equations. Conductivity. Some materials consist of an abundance of electrons which are loosely-bound to the atoms and molecules comprising the material. The force exerted on these electrons by an electric field may be sufficient to overcome the binding force, resulting in motion of the associated charges and subsequently current. This effect is quantified by Ohm’s law for electromagnetics: J = σE

(1.15)

where J in this case is the conduction current B = µH (1.11) determined by the conductivity σ (SI base units of S/m). Conductivity is a property of a material that where µ is the permeability (SI base units of H/m) of ranges from negligible (i.e., for “insulators”) to very large for good conductors, which includes most the material. In free space, µ is equal to metals. µ0 , 4π × 10−7 H/m. (1.12) A perfect conductor is a material within which E is It is often convenient to quantify the permeability of essentially zero regardless of J. For such material, material in terms of the unitless relative permeability σ → ∞. Perfect conductors are said to be µr , µ/µ0 . equipotential regions; that is, the potential difference

1.4. ELECTROMAGNETIC FIELD THEORY: A REVIEW

Electric & magnetic fields are... Maxwell’s eqns. (integral)

Maxwell’s eqns. (differential)

7

Electrostatics / Magnetostatics independent

Time-Varying (Dynamic) possibly coupled

H HS D · ds = Qencl HC E · dl = 0 HS B · ds = 0 H · dl = Iencl C ∇ · D = ρv ∇×E=0 ∇·B=0 ∇×H=J

H HS D · ds = Q∂encl R HC E · dl = − ∂t S B · ds HS B · ds = 0 R ∂ H · dl = Iencl + S ∂t D · ds C ∇ · D = ρv ∂ ∇ × E = − ∂t B ∇·B=0 ∂ ∇ × H = J+ ∂t D

Table 1.3: Comparison of Maxwell’s equations for static and time-varying electromagnetic fields. Differences in the time-varying case relative to the static case are highlighted in blue. between any two points within a perfect conductor is zero, as can be readily verified using Equation 1.1. Time-varying fields. Faraday’s law states that a time-varying magnetic flux induces an electric potential in a closed loop as follows: V =−

∂ Φ ∂t

(1.16)

Setting this equal to the left side of Equation 1.2 leads to the Maxwell-Faraday equation in integral form: I Z ∂ B · ds (1.17) E · dl = − ∂t S C where C is the closed path defined by the edge of the open surface S. Thus, we see that a time-varying magnetic flux is able to generate an electric field. We also observe that electric and magnetic fields become coupled when the magnetic flux is time-varying. An analogous finding leads to the general form of Ampere’s law: I Z ∂ H · dl = Iencl + D · ds (1.18) C S ∂t

Gauss’ law for electric and magnetic fields, boundary conditions, and constitutive relationships (Equations 1.5, 1.11, and 1.15) are the same in the time-varying case. As indicated in Table 1.3, the time-varying version of Maxwell’s equations may also be expressed in differential form. The differential forms make clear that variations in the electric field with respect to position are associated with variations in the magnetic field with respect to time (the Maxwell-Faraday equation), and vice-versa (Ampere’s law). Time-harmonic waves in source-free and lossless media. The coupling between electric and magnetic fields in the time-varying case leads to wave phenomena. This is most easily analyzed for fields which vary sinusoidally, and may thereby be expressed as phasors.1 Phasors, indicated in this book by the tilde (“e”), are complex-valued quantities representing the magnitude and phase of the associated sinusoidal waveform. Maxwell’s equations in differential phasor form are: e = ρev ∇·D e = −jω B e ∇×E

e =0 ∇·B e =J e + jω D e ∇×H

(1.19) (1.20)

(1.21) where the new term is referred to as displacement current. Through the displacement current, a (1.22) time-varying electric flux may be a source of the magnetic field. In other words, we see that the electric where ω , 2πf , and where f is frequency (SI base 1 Sinusoidally-varying fields are sometimes also said to be timeand magnetic fields are coupled when the electric flux harmonic. is time-varying.

CHAPTER 1. PRELIMINARY CONCEPTS

8 units of Hz). In regions which are free of sources (i.e., charges and currents) and consisting of loss-free media (i.e., σ = 0), these equations reduce to the following: e =0 ∇·E e = −jωµH e ∇×E

e =0 ∇·H e = +jωǫE e ∇×H

The scalar components of the plane waves described in Equations 1.31 and 1.32 exhibit the same characteristics as other types of waves, including sound waves and voltage and current waves in transmission lines. In particular, the phase velocity of (1.23) waves propagating in the +ˆ z and −ˆ z direction is (1.24)

vp =

(1.25)

where

√ β , ω µǫ

(1.33)

2π β

(1.34)

(1.26) and the wavelength is

where we have used the relationships D = ǫE and B = µH to eliminate the flux densities D and B, which are now redundant. Solving Equations 1.23–1.26 for E and H, we obtain the vector wave equations: e + β2E e =0 ∇2 E e + β2H e =0 ∇2 H

1 ω =√ β µǫ

λ=

e and H e to satisfy the By requiring solutions for E Maxwell curl equations (i.e., the Maxwell-Faraday e H, e and equation and Ampere’s law), we find that E, (1.27) the direction of propagation k ˆ are mutually (1.28) perpendicular. In particular, we obtain the plane wave relationships: (1.29)

Waves in source-free and lossless media are solutions to the vector wave equations.

where

ˆ×H e = −η k e E ˆ×E e e = 1k H η

(1.35) (1.36)

r

Uniform plane waves in source-free and lossless media. An important subset of solutions to the vector wave equations are uniform plane waves. Uniform plane waves result when solutions are constrained to exhibit constant magnitude and phase in a plane. For example, if this plane is specified to be perpendicular e to z (i.e., ∂/∂x = ∂/∂y = 0) then solutions for E have the form:

µ (1.37) ǫ is the wave impedance, also known as the intrinsic ˆ is in the same impedance of the medium, and k e e direction as E × H.

where

where S has SI base units of W/m2 , and here it is e is in peak (as opposed to rms) units. assumed that E

e =x ex + y ey ˆE ˆE E

ex = E + e−jβz + E − e+jβz E x0 x0 + −jβz − +jβz e Ey = Ey0 e + Ey0 e

η,

The power density associated with a plane wave is 2 e E S= (1.38) (1.30) 2η

(1.31)

(1.32) Commonly-assumed properties of materials. Finally, a reminder about commonly-assumed + − + − properties of the material constitutive parameters ǫ, µ, and where Ex0 , Ex0 , Ey0 , and Ey0 are constant complex-valued coefficients which depend on sources and σ. We often assume these parameters exhibit the following properties: and boundary conditions. The first term and second terms of Equations 1.31 and 1.32 correspond to waves traveling in the +ˆ z and −ˆ z directions, respectively. e is a solution to the same vector wave Because H equation, the solution for H is identical except with different coefficients.

• Homogeneity. A material that is homogeneous is uniform over the space it occupies; that is, the values of its constitutive parameters are constant at all locations within the material.

1.4. ELECTROMAGNETIC FIELD THEORY: A REVIEW • Isotropy. A material that is isotropic behaves in precisely the same way regardless of how it is oriented with respect to sources, fields, and other materials. • Linearity. A material is said to be linear if its properties do not depend on the sources and fields applied to the material. Linear media exhibit superposition; that is, the response to multiple sources is equal to the sum of the responses to the sources individually. • Time-invariance. A material is said to be time-invariant if its properties do not vary as a function of time. Additional Reading: • “Maxwell’s Equations” on Wikipedia. • “Wave Equation” on Wikipedia. • “Electromagnetic Wave Equation” on Wikipedia. • “Electromagnetic radiation” on Wikipedia.

[m0181]

9

10

CHAPTER 1. PRELIMINARY CONCEPTS

Image Credits c K. Kikkeri, https://commons.wikimedia.org/wiki/File:M0006 fCartesianBasis.svg, Fig. 1.1: CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c K. Kikkeri, https://commons.wikimedia.org/wiki/File:M0096 fCylindricalCoordinates.svg, Fig. 1.2: CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c K. Kikkeri, https://commons.wikimedia.org/wiki/File:Spherical Coordinate System.svg, Fig. 1.3: CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/).

Chapter 2

Magnetostatics Redux 2.1 Lorentz Force Additional Reading: [m0015]

The Lorentz force is the force experienced by charge in the presence of electric and magnetic fields.

• “Lorentz force” on Wikipedia.

Consider a particle having charge q. The force Fe experienced by the particle in the presence of electric field intensity E is Fe = qE The force Fm experienced by the particle in the presence of magnetic flux density B is Fm = qv × B where v is the velocity of the particle. The Lorentz force experienced by the particle is simply the sum of these forces; i.e., F = Fe + Fm = q (E + v × B)

(2.1)

The term “Lorentz force” is simply a concise way to refer to the combined contributions of the electric and magnetic fields. c Stannered CC BY 2.5.

A common application of the Lorentz force concept is in analysis of the motions of charged particles in electromagnetic fields. The Lorentz force causes charged particles to exhibit distinct rotational (“cyclotron”) and translational (“drift”) motions. This is illustrated in Figures 2.1 and 2.2.

Figure 2.1: Motion of a particle bearing (left) positive charge and (right) negative charge. Top: Magnetic field directed toward the viewer; no electric field. Bottom: Magnetic field directed toward the viewer; electric field oriented as shown.

c 2020 S.W. Ellingson CC BY SA 4.0. https://doi.org/10.21061/electromagnetics-vol-2 Electromagnetics Vol. 2.

CHAPTER 2. MAGNETOSTATICS REDUX

12

Let dl (r) represent a differential-length segment of the wire at r, pointing in the direction of current flow. Then dq v (r) = Idl (r) (2.4) (If this is not clear, it might help to consider the units: On the left, C·m/s = (C/s)·m = A·m, as on the right.) Subsequently,

c M. Biaek CC BY-SA 4.0.

Figure 2.2: Electrons moving in a circle in a magnetic field (cyclotron motion). The electrons are produced by an electron gun at bottom, consisting of a hot cathode, a metal plate heated by a filament so it emits electrons, and a metal anode at a high voltage with a hole which accelerates the electrons into a beam. The electrons are normally invisible, but enough air has been left in the tube so that the air molecules glow pink when struck by the fast-moving electrons.

2.2 Magnetic Force on a Current-Carrying Wire

dFm (r) = Idl (r) × B (r)

(2.5)

There are three important cases of practical interest. First, consider a straight segment l forming part of a closed loop of current in a spatially-uniform impressed magnetic flux density B (r) = B0 . In this case, the force exerted by the magnetic field on such a segment is given by Equation 2.5 with dl replaced by l; i.e.: (2.6) Fm = Il × B0 Summarizing, The force experienced by a straight segment of current-carrying wire in a spatially-uniform magnetic field is given by Equation 2.6.

[m0017]

The second case of practical interest is a rigid closed loop of current in a spatially-uniform magnetic flux Consider an infinitesimally-thin and density B0 . If the loop consists of straight sides – perfectly-conducting wire bearing a current I (SI base e.g., a rectangular loop – then the force applied to the units of A) in free space. Let B (r) be the impressed loop is the sum of the forces applied to each side magnetic flux density at each point r in the region of separately, as determined by Equation 2.6. However, space occupied by the wire. By impressed, we mean we wish to consider loops of arbitrary shape. To that the field exists in the absence of the accommodate arbitrarily-shaped loops, let C be the current-carrying wire, as opposed to the field that is path through space occupied by the loop. Then the induced by this current. Since current consists of force experienced by the loop is charged particles in motion, we expect that B(r) will Z exert a force on the current. Since the current is F = dFm (r) constrained to flow on the wire, we expect this force C Z will also be experienced by the wire. Let us now = Idl (r) × B0 (2.7) consider this force. C

To begin, recall that the force exerted on a particle bearing charge q having velocity v is

Since I and B0 are constants, they may be extracted from the integral: Z  Fm (r) = qv (r) × B (r) (2.2) F=I dl (r) × B0 (2.8) Thus, the force exerted on a differential amount of C charge dq is Note the quantity in square brackets is zero. dFm (r) = dq v (r) × B (r) (2.3) Therefore:

2.2. MAGNETIC FORCE ON A CURRENT-CARRYING WIRE The net force on a current-carrying loop of wire in a uniform magnetic field is zero.

Note that this does not preclude the possibility that the rigid loop rotates; for example, the force on opposite sides of the loop may be equal and opposite. What we have found is merely that the force will not lead to a translational net force on the loop; e.g., force that would propel the loop away from its current position in space. The possibility of rotation without translation leads to the most rudimentary concept for an electric motor. Practical electric motors use variations on essentially this same idea; see “Additional Reading” for more information. The third case of practical interest is the force experienced by two parallel infinitesimally-thin wires in free space, as shown in Figure 2.3. Here the wires are infinite in length (we’ll return to that in a moment), lie in the x = 0 plane, are separated by distance d, and carry currents I1 and I2 , respectively. The current in wire 1 gives rise to a magnetic flux density B1 . The force exerted on wire 2 by B1 is: F2 =

Z

C

[I2 dl (r) × B1 (r)]

(2.9)

13

Returning to Equation 2.9, we obtain:   Z  µ0 I 1 ˆdz × −ˆ F2 = I2 z x 2πd C Z µ0 I 1 I 2 = −ˆ y dz 2πd C

(2.11)

The remaining integral is simply the length of wire 2 that we wish to consider. Infinitely-long wires will therefore result in infinite force. This is not a very interesting or useful result. However, the force per unit length of wire is finite, and is obtained simply by dropping the integral in the previous equation. We obtain: µ0 I 1 I 2 F2 = −ˆ y (2.12) ∆l 2πd where ∆l is the length of the section of wire 2 being considered. Note that when the currents I1 and I2 flow in the same direction (i.e., have the same sign), the magnetic force exerted by the current on wire 1 pulls wire 2 toward wire 1. The same process can be used to determine the magnetic force F1 exerted by the current in wire 1 on wire 2. The result is F1 µ0 I 1 I 2 = +ˆ y (2.13) ∆l 2πd

ˆdz. where C is the path followed by I2 , and dl (r) = z A simple way to determine B1 in this situation is as follows. First, if wire 1 had been aligned along the x = y = 0 line, then the magnetic flux density everywhere would be µ0 I 1 φˆ 2πρ In the present problem, wire 1 is displaced by d/2 in the −ˆ y direction. Although this would seem to make the new expression more complicated, note that the only positions where values of B1 (r) are required are those corresponding to C; i.e., points on wire 2. For these points, B1 (r) = −ˆ x

µ0 I 1 along C 2πd

(2.10) c Y. Zhao CC BY-SA 4.0

That is, the relevant distance is d (not ρ), and the ˆ direction of B1 (r) for points along C is −ˆ x (not φ).

Figure 2.3: Parallel current-carrying wires.

CHAPTER 2. MAGNETOSTATICS REDUX

14 When the currents I1 and I2 flow in the same direction (i.e., when the product I1 I2 is positive), then the magnetic force exerted by the current on wire 2 pulls wire 1 toward wire 2. We are now able to summarize the results as follows: If currents in parallel wires flow in the same direction, then the wires attract; whereas if the currents flow in opposite directions, then the wires repel.

with the negative sign indicating that the wires repel. Note in the above example that this force is quite small, which explains why it is not always observed. However, this force becomes significant when the current is large or when many sets of conductors are mechanically bound together (amounting to a larger net current), as in a motor. Additional Reading:

Also: The magnitude of the associated force is µ0 I1 I2 /2πd for wires separated by distance d in non-magnetic media. If the wires are fixed in position and not able to move, these forces represent stored (potential) energy. It is worth noting that this is precisely the energy which is stored by an inductor – for example, the two wire segments here might be interpreted as segments in adjacent windings of a coil-shaped inductor. Example 2.1. DC power cable. A power cable connects a 12 V battery to a load exhibiting an impedance of 10 Ω. The conductors are separated by 3 mm by a plastic insulating jacket. Estimate the force between the conductors. Solution. The current flowing in each conductor is 12 V divided by 10 Ω, which is 1.2 A. In terms of the theory developed in this section, a current I1 = +1.2 A flows from the positive terminal of the battery to the load on one conductor, and a current I2 = −1.2 A returns to the battery on the other conductor. The change in sign indicates that the currents at any given distance from the battery are flowing in opposite directions. Also from the problem statement, d = 3 mm and the insulator is presumably non-magnetic. Assuming the conductors are approximately straight, the force between conductors is ≈

µ0 I 1 I 2 ∼ = −96.0 µN 2πd

• “Electric motor” on Wikipedia.

2.3. TORQUE INDUCED BY A MAGNETIC FIELD

2.3 Torque Induced by a Magnetic Field

which points in a direction perpendicular to both d and F.

[m0024]

A magnetic field exerts a force on current. This force is exerted in a direction perpendicular to the direction of current flow. For this reason, current-carrying structures in a magnetic field tend to rotate. A convenient description of force associated with rotational motion is torque. In this section, we define torque and apply this concept to a closed loop of current. These concepts apply to a wide range of practical devices, including electric motors. Figure 2.4 illustrates the concept of torque. Torque depends on the following: • A local origin r0 ,

Note that T does not point in the direction of rotation. Nevertheless, T indicates the direction of rotation through a “right hand rule”: If you point the thumb of your right hand in the direction of T, then the curled fingers of your right hand will point in the direction of torque-induced rotation. Whether rotation actually occurs depends on the geometry of the structure. For example, if T aligns with the axis of a perfectly-rigid mechanical shaft, then all of the work done by F will be applied to rotation of the shaft on this axis. Otherwise, torque will tend to rotate the shaft in other directions as well. If the shaft is not free to rotate in these other directions, then the effective torque – that is, the torque that contributes to rotation of the shaft – is reduced. The magnitude of T has SI base units of N·m and quantifies the energy associated with the rotational force. As you might expect, the magnitude of the torque increases with increasing lever arm magnitude |d|. In other words, the torque resulting from a constant applied force increases with the length of the lever arm.

• A point r which is connected to r0 by a perfectly-rigid mechanical structure, and • The force F applied at r. In terms of these parameters, the torque T is: T,d×F

15

(2.14)

where the lever arm d , r − r0 gives the location of r relative to r0 . Note that T is a position-free vector

Torque, like the translational force F, satisfies superposition. That is, the torque resulting from forces applied to multiple rigidly-connected lever arms is the sum of the torques applied to the lever arms individually. Now consider the current loop shown in Figure 2.5. The loop is perfectly rigid and is rigidly attached to a non-conducting shaft. The assembly consisting of the loop and the shaft may rotate without friction around the axis of the shaft. The loop consists of four straight segments that are perfectly-conducting and infinitesimally-thin. A spatially-uniform and static ˆ B0 exists impressed magnetic flux density B0 = x throughout the domain of the problem. (Recall that an impressed field is one that exists in the absence of any other structure in the problem.) What motion, if any, is expected?

c C. Wang CC BY-SA 4.0

Figure 2.4: Torque associated with a single lever arm.

Recall that the net translational force on a current loop in a spatially-uniform and static magnetic field is

CHAPTER 2. MAGNETOSTATICS REDUX

16 zero (Section 2.2). However, this does not preclude the possibility of different translational forces acting on each of the loop segments resulting in a rotation of the shaft. Let us first calculate these forces. The force FA on segment A is FA = IlA × B0

We calculate the associated torque T as T = TA + TB + TC + TD

where TA , TB , TC , and TD are the torques associated with segments A, B, C, and D, respectively. (2.15) For example, the torque associated with segment A is

where lA is a vector whose magnitude is equal to the length of the segment and which points in the direction of the current. Thus, FA = I (ˆ zL) × (ˆ xB0 ) ˆ ILB0 =y

TB = 0 since FB = 0 W IB0 2 = 0 since FD = 0

ˆL TC = z (2.17)

(2.21)

TD

(2.22) (2.23) (2.24)

Summing these contributions, we find

The forces FB and FD on segments B and D, respectively, are: FB = I (−ˆ xL) × (ˆ xB0 ) = 0

W ˆ × FA x 2 W ˆL IB0 =z 2

TA =

(2.16) Similarly,

Similarly, the force FC on segment C is FC = I (−ˆ zL) × (ˆ xB0 ) = −ˆ yILB0

(2.20)

ˆLW IB0 T=z

(2.25)

(2.18)

Note that T points in the +ˆ z direction, indicating rotational force exerted in the +φˆ direction, as FD = I (+ˆ xL) × (ˆ xB0 ) = 0 (2.19) expected. Also note that the torque is proportional to the area LW of the loop, is proportional to the current Thus, the force exerted on the current loop by the I, and is proportional to the magnetic field magnitude impressed magnetic field will lead to rotation in the ˆ B0 . +φ direction. and

The analysis that we just completed was static; that is, it applies only at the instant depicted in Figure 2.5. If B0=x̂ B0 the shaft is allowed to turn without friction, then the B loop will rotate in the +φˆ direction. So, what will loop wire L happen to the forces and torque? First, note that FA and FC are always in the +ˆ y and −ˆ y directions, respectively, regardless of the rotation of the loop. C A Once the loop rotates away from the position shown x in Figure 2.5, the forces FB and FD become non-zero; however, they are always equal and opposite, and so do not affect the rotation. Thus, the I loop will rotate one-quarter turn and then come to current rest, perhaps with some damped oscillation around D source the rest position depending on the momentum of the W loop. At the rest position, the lever arms for non-conducting shaft segments A and C are pointing in the same directions as FA and FC , respectively. Therefore, the cross c C. Wang CC BY-SA 4.0

product of the lever arm and translational force for each segment is zero and subsequently Figure 2.5: A rudimentary electric motor consisting TA = TC = 0. Once stopped in this position, both of a single current loop. the net translational force and the net torque are zero. z

2.3. TORQUE INDUCED BY A MAGNETIC FIELD

17 Additional Reading: • “Torque” on Wikipedia. • “Electric motor” on Wikipedia.

c Abnormaal CC BY-SA 3.0

Figure 2.6: This DC electric motor uses brushes (here, the motionless leads labeled “+” and “−”) combined with the motion of the shaft to periodically alternate the direction of current between two coils, thereby creating nearly constant torque.

If such a device is to be used as a motor, it is necessary to find a way to sustain the rotation. There are several ways in which this might be accomplished. First, one might make I variable in time. For example, the direction of I could be reversed as the loop passes the quarter-turn position. This reverses FA and FC , propelling the loop toward the half-turn position. The direction of I can be changed again as the loop passes half-turn position, propelling the loop toward the three-quarter-turn position. Continuing this periodic reversal of the current sustains the rotation. Alternatively, one may periodically reverse the direction of the impressed magnetic field to the same effect. These methods can be combined or augmented using multiple current loops or multiple sets of time-varying impressed magnetic fields. Using an appropriate combination of current loops, magnetic fields, and waveforms for each, it is possible to achieve sustained torque throughout the rotation. An example is shown in Figure 2.6.

CHAPTER 2. MAGNETOSTATICS REDUX

18

2.4 The Biot-Savart Law In addition to obviating the need to solve a differential equation, BSL provides some useful insight into the behavior of magnetic fields. In particular, The Biot-Savart law (BSL) provides a method to Equation 2.28 indicates that magnetic fields follow calculate the magnetic field due to any distribution of the inverse square law – that is, the magnitude of the steady (DC) current. In magnetostatics, the general magnetic field due to a differential current element solution to this problem employs Ampere’s law; i.e., decreases in proportion to the inverse square of Z distance (R−2 ). Also, Equation 2.28 indicates that the H · dl = Iencl (2.26) direction of the magnetic field due to a differential C current element is perpendicular to both the direction in integral form or ˆ pointing from the of current flow ˆl and the vector R ∇×H=J (2.27) source point to field point. This observation is quite useful in anticipating the direction of magnetic field in differential form. The integral form is relatively vectors in complex problems. simple when the problem exhibits a high degree of It may be helpful to note that BSL is analogous to symmetry, facilitating a simple description in a Coulomb’s law for electric fields, which is a solution particular coordinate system. An example is the to the differential form of Gauss’ law, ∇ · D = ρv . magnetic field due to a straight and infinitely-long However, BSL applies only under magnetostatic current filament, which is easily determined by conditions. If the variation in currents or magnetic solving the integral equation in cylindrical fields over time is significant, then the problem coordinates. However, many problems of practical becomes significantly more complicated. See interest do not exhibit the necessary symmetry. A commonly-encountered example is the magnetic field “Jefimenko’s Equations” in “Additional Reading” for more information. due to a single loop of current, which will be addressed in Example 2.2. For such problems, the differential form of Ampere’s law is needed. Example 2.2. Magnetic field along the axis of a circular loop of current. BSL is the solution to the differential form of Ampere’s law for a differential-length current Consider a ring of radius a in the z = 0 plane, element, illustrated in Figure 2.7. The current element centered on the origin, as shown in Figure 2.8. is I dl, where I is the magnitude of the current (SI As indicated in the figure, the current I flows in base units of A) and dl is a differential-length vector indicating the direction of the current at the “source point” r′ . The resulting contribution to the magnetic field intensity at the “field point” r is dH(r) [m0066]

dH(r) = I dl where

1 ˆ ×R 4πR2

ˆ , r − r′ R = RR

(2.28) RR

(2.29)

dl @r'

In other words, R is the vector pointing from the I source point to the field point, and dH at the field point is given by Equation 2.28. The magnetic field due to a current-carrying wire of any shape may be c C. Wang CC BY-SA 4.0

obtained by integrating over the length of the wire: Z Z Figure 2.7: Use of the Biot-Savart law to calculate the ˆ I dl × R H(r) = dH(r) = (2.30) magnetic field due to a line current. 4π C R2 C

2.4. THE BIOT-SAVART LAW

19

the φˆ direction. Find the magnetic field intensity along the z axis. Solution. The source current position is given in cylindrical coordinates as r′ = ρˆa

(2.31)

The position of a field point along the z axis is ˆz r=z

(2.32)

Thus,

c K. Kikkeri CC BY SA 4.0 (modified)

ˆ , r − r′ = −ˆ ˆz RR ρa + z

(2.33)

and R , |r − r′ | =

p

a2 + z 2

(2.34) The first integral in the previous equation is equal to 2π. Thus, we obtain

Equation 2.28 becomes: ˆz − ρˆa I φˆ adφ z dH(ˆ zz) = ×√ 2 2 4π [a + z ] a2 + z 2 ˆa − ρˆz Ia z dφ (2.35) = 4π [a2 + z 2 ]3/2 Now integrating over the current: H(ˆ zz) =

=

=

Z

2π 0

ˆa − ρˆz Ia z dφ 4π [a2 + z 2 ]3/2

Ia 4π [a2 + z 2 ]

3/2

Ia 4π [a2 + z 2 ]

3/2

Z

Figure 2.8: Calculation of the magnetic field along the z axis due to a circular loop of current centered in the z = 0 plane.

(2.36)

ˆ H(ˆ zz) = z

Ia2 2 [a2 + z 2 ]

3/2

(2.39)

Note that the result is consistent with the associated “right hand rule” of magnetostatics: That is, the direction of the magnetic field is in the direction of the curled fingers of the right hand when the thumb of the right hand is aligned with the location and direction of current. It is a good exercise to confirm that this result is also dimensionally correct.

2π 0



(ˆ za − ρˆz) dφ (2.37)

ˆa z

Z

2π 0

dφ − z

Z



ρˆ dφ 0



(2.38)

The second integral is equal to zero. To see this, note that the integral is simply summing values of ρˆ for all possible values of φ. Since ρˆ(φ + π) = −ˆ ρ(φ), the integrand for any given value of φ is equal and opposite the integrand π radians later. (This is one example of a symmetry argument.)

Equation 2.28 extends straightforwardly to other distributions of current. For example, the magnetic field due to surface current Js (SI base units of A/m) can be calculated using Equation 2.28 with I dl replaced by Js ds where ds is the differential element of surface area. This can be confirmed by dimensional analysis: I dl has SI base units of A·m, as does JS ds. Similarly, the magnetic field due to volume current J (SI base units of A/m2 ) can be calculated using Equation 2.28 with I dl replaced by J dv where dv is the differential element of volume. For a single particle with charge q (SI base units of C) and

CHAPTER 2. MAGNETOSTATICS REDUX

20 velocity v (SI base units of m/s), the relevant quantity is qv since C·m/s = (C/s)·m = A·m. In all of these cases, Equation 2.28 applies with the appropriate replacement for I dl. Note that the quantities qv, I dl, JS ds, and J dv, all having the same units of A·m, seem to be referring to the same physical quantity. This physical quantity is known as current moment. Thus, the “input” to BSL can be interpreted as current moment, regardless of whether the current of interest is distributed as a line current, a surface current, a volumetric current, or simply as moving charged particles. See “Additional Reading” at the end of this section for additional information on the concept of “moment” in classical physics. Additional Reading: • “Biot-Savart Law” on Wikipedia. • “Jefimenko’s Equations” on Wikipedia. • “Moment (physics)” on Wikipedia.

2.5 Force, Energy, and Potential Difference in a Magnetic Field [m0059]

The force Fm experienced by a particle at location r bearing charge q due to a magnetic field is Fm = qv × B(r)

(2.40)

where v is the velocity (magnitude and direction) of the particle, and B(r) is the magnetic flux density at r. Now we must be careful: In this description, the motion of the particle is not due to Fm . In fact the cross product in Equation 2.40 clearly indicates that Fm and v must be in perpendicular directions. Instead, the reverse is true: i.e., it is the motion of the particle that is giving rise to the force. The motion described by v may be due to the presence of an electric field, or it may simply be that that charge is contained within a structure that is itself in motion. Nevertheless, the force Fm has an associated potential energy. Furthermore, this potential energy may change as the particle moves. This change in potential energy may give rise to an electrical potential difference (i.e., a “voltage”), as we shall now demonstrate. The change in potential energy can be quantified using the concept of work, W . The incremental work ∆W done by moving the particle a short distance ∆l, over which we assume the change in Fm is negligible, is ∆W ≈ Fm · ˆl∆l (2.41) where in this case ˆl is the unit vector in the direction of the motion; i.e., the direction of v. Note that the purpose of the dot product in Equation 2.41 is to ensure that only the component of Fm parallel to the direction of motion is included in the energy tally. Any component of v which is due to Fm (i.e., ultimately due to B) must be perpendicular to Fm , so ∆W for such a contribution must be, from Equation 2.41, equal to zero. In other words: In the absence of a mechanical force or an electric field, the potential energy of a charged particle remains constant regardless of how it is moved by Fm . This surprising result may be summarized as follows:

2.5. FORCE, ENERGY, AND POTENTIAL DIFFERENCE IN A MAGNETIC FIELD

21

At this point, it is convenient to introduce the electric potential difference V21 between the start point (1) and end point (2) of C. V21 is defined as the work done by traversing C, per unit of charge; i.e.,

The magnetic field does no work.

Instead, the change of potential energy associated with the magnetic field must be completely due to a W change in position resulting from other forces, such as V21 , (2.47) q a mechanical force or the Coulomb force. The presence of a magnetic field merely increases or This has units of J/C, which is volts (V). Substituting decreases this potential difference once the particle Equation 2.46, we obtain: has moved, and it is this change in the potential difference that we wish to determine. Z (2.48) V21 = [v × B] · dl We can make the relationship between potential C difference and the magnetic field explicit by substituting the right side of Equation 2.40 into Equation 2.41, yielding Equation 2.48 is electrical potential induced by charge traversing a magnetic field. ∆W ≈ q [v × B(r)] · ˆl∆l (2.42) Equation 2.42 gives the work only for a short distance around r. Now let us try to generalize this result. If we wish to know the work done over a larger distance, then we must account for the possibility that v × B varies along the path taken. To do this, we may sum contributions from points along the path traced out by the particle, i.e., W ≈

N X

∆W (rn )

Figure 2.9 shows a simple scenario that illustrates this concept. Here, a straight perfectly-conducting wire of length l is parallel to the y axis and moves at speed v ˆ B. in the +z direction through a magnetic field B = x Thus, ˆv × x ˆB v×B=z ˆ Bv =y

Taking endpoints 1 and 2 of the wire to be at y = y0

n=1

where rn are positions defining the path. Substituting the right side of Equation 2.42, we have W ≈q

N X

n=1

[v × B(rn )] · ˆl(rn )∆l

Taking the limit as ∆l → 0, we obtain Z W = q [v × B(r)] · ˆl(r)dl

(2.49)

(2.43)

(2)

B

(2.44)

(2.45)

v

l

C

where C is the path (previously, the sequence of rn ’s) followed by the particle. Now omitting the explicit dependence on r in the integrand for clarity: Z W = q [v × B] · dl (2.46) C

where dl = ˆldl as usual. Now, we are able to determine the change in potential energy for a charged particle moving along any path in space, given the magnetic field.

y (1) x

z

c C. Wang CC BY 4.0

Figure 2.9: A straight wire moving through a magnetic field.

CHAPTER 2. MAGNETOSTATICS REDUX

22

y

and y = y0 + l, respectively, we obtain Z y0 +l ˆ dy V21 = [ˆ yBv] · y y0

= Bvl

(2.50)

+ VT

R

B

v

Thus, we see that endpoint 2 is at an electrical potential of Bvl greater than that of endpoint 1. This “voltage” exists even though the wire is perfectly-conducting, and therefore cannot be attributed to the electric field. This voltage exists even x z though the force required for movement must be the c C. Wang CC BY-SA 4.0

same on both endpoints, or could even be zero, and therefore cannot be attributed to mechanical forces. Figure 2.10: A time-varying loop created by moving Instead, this change in potential is due entirely to the a “shorting bar” along rails comprising two adjacent magnetic field. sides of the loop. Because the wire does not form a closed loop, no current flows in the wire. Therefore, this scenario has ˆv (constant). What is the voltage VT v=z limited application in practice. To accomplish across the resistor and what is the current in the something useful with this concept we must at least loop? form a closed loop, so that current may flow. For a closed loop, Equation 2.48 becomes: Solution. Since the gap containing the resistor is I infinitesimally small, V = [v × B] · dl (2.51) I C VT = [v × B] · dl (2.52) Examination of this equation indicates one additional

requirement: v × B must somehow vary over C. This is because if v × B does not vary over C, the result will be I [v × B] · dl C

which is zero because the integral is zero. The following example demonstrates a practical application of this idea. Example 2.3. Potential induced in a time-varying loop.

Figure 2.10 shows a modification to the problem originally considered in Figure 2.9. Now, we have created a closed loop using perfectly-conducting and motionless wire to form three sides of a rectangle, and assigned the origin to the lower left corner. An infinitesimally-small gap has been inserted in the left (z = 0) side of the loop and closed with an ˆB ideal resistor of value R. As before, B = x (spatially uniform and time invariant) and

C

where C is the perimeter formed by the loop, beginning at the “−” terminal of VT and returning to the “+” terminal of VT . Only the shorting bar is in motion, so v = 0 for the other three sides of the loop. Therefore, only the portion of C traversing the shorting bar contributes to VT . Thus, we find Z l ˆ B] · y ˆ dy = Bvl (2.53) VT = [ˆ zv × x y=0

This potential gives rise to a current Bvl/R, which flows in the counter-clockwise direction.

Note in the previous example that the magnetic field has induced VT , not the current. The current is simply a response to the existence of the potential, regardless of the source. In other words, the same potential VT would exist even if the gap was not closed by a resistor.

2.5. FORCE, ENERGY, AND POTENTIAL DIFFERENCE IN A MAGNETIC FIELD Astute readers will notice that this analysis seems to have a lot in common with Faraday’s law, V =−

∂ Φ ∂t

(2.54)

which says the potential induced in a single closed loop is proportional to the time rate of change of magnetic flux Φ, where Z Φ= B · ds (2.55) S

and where S is the surface through which the flux is calculated. From this perspective, we see that Equation 2.51 is simply a special case of Faraday’s law, pertaining specifically to “motional emf.” Thus, the preceding example can also be solved by Faraday’s law, taking S to be the time-varying surface bounded by C.

[m0182]

23

24

CHAPTER 2. MAGNETOSTATICS REDUX

Image Credits c Stannered, https://en.wikipedia.org/wiki/File:Charged-particle-drifts.svg, Fig. 2.1: CC BY 2.5 (https://creativecommons.org/licenses/by/2.5/deed.en). Modified by Maschen, author. c M. Biaek, https://en.wikipedia.org/wiki/File:Cyclotron motion.jpg, Fig. 2.2: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c YahuiZ (Y. Zhao), https://commons.wikimedia.org/wiki/File:Figure2.3Yahui.svg, Fig. 2.3: CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). Modified by author. c Sevenchw (C. Wang), Fig. 2.4: https://commons.wikimedia.org/wiki/File:Torque associated with a single lever arm.svg, CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), https://commons.wikimedia.org/wiki/File:Rudimentary electric motor.svg, Fig. 2.5: CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Abnormaal, https://en.wikipedia.org/wiki/File:Electric motor.gif, Fig. 2.6: CC BY SA 3.0 (https://creativecommons.org/licenses/by-sa/3.0/). c Sevenchw (C. Wang), Fig. 2.7: https://commons.wikimedia.org/wiki/File:Magnetic field due to a line current.svg, CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c K. Kikkeri, https://commons.wikimedia.org/wiki/File:M0104 fRing.svg, Fig. 2.8: CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). Modified by author. c Sevenchw (C. Wang), Fig. 2.9: https://commons.wikimedia.org/wiki/File:Straight wire moving in a magnetic field.svg, CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 2.10: https://commons.wikimedia.org/wiki/File:Moving shorting bar of a circuit in mag field.svg, CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/).

Chapter 3

Wave Propagation in General Media 3.1 Poynting’s Theorem

storage in magnetic fields within V, and PΩ is the power dissipated (converted to heat) in V.

[m0073]

Some preliminary mathematical results. We now derive two mathematical relationships that will be useful later. Let E be the electric field intensity (SI base units of V/m) and let D be the electric flux density (SI base units of C/m2 ). We require that the constituents of V be linear and time-invariant; therefore, D = ǫE where the permittivity ǫ is constant with respect to time, but not necessarily with respect to position. Under these conditions, we find

Despite the apparent complexity of electromagnetic theory, there are in fact merely four ways that electromagnetic energy can be manipulated. Electromagnetic energy can be: • Transferred; i.e., conveyed by transmission lines or in waves; • Stored in an electric field (capacitance);

1 ∂ ∂ (E · D) = (D · D) ∂t ǫ ∂t

• Stored in a magnetic field (inductance); or • Dissipated (converted to heat; i.e., resistance). Poynting’s theorem is an expression of conservation of energy that elegantly relates these various possibilities. Once recognized, the theorem has important applications in the analysis and design of electromagnetic systems. Some of these emerge from the derivation of the theorem, as opposed to the unsurprising result. So, let us now derive the theorem.

(3.2)

ˆ Note that E and D point in the same direction. Let e be the unit vector describing this direction. Then, ˆE and D = e ˆD where E and D are the scalar E=e

We begin with a statement of the theorem. Consider a volume V that may contain materials and structures in any combination. This is crudely depicted in Figure 3.1. Also recall that power is the time rate of change of energy. Then: Pnet,in = PE + PM + PΩ

(3.1)

where Pnet,in is the net power flow into V, PE is the power associated with energy storage in electric fields within V, PM is the power associated with energy

c C. Wang CC BY-SA 4.0

Figure 3.1: Poynting’s theorem describes the fate of power entering a region V consisting of materials and structures capable of storing and dissipating energy.

c 2020 S.W. Ellingson CC BY SA 4.0. https://doi.org/10.21061/electromagnetics-vol-2 Electromagnetics Vol. 2.

CHAPTER 3. WAVE PROPAGATION IN GENERAL MEDIA

26

components of E and D, respectively. Subsequently, 1 ∂ 1 ∂ 2 (D · D) = D ǫ ∂t ǫ ∂t 1 ∂ = 2D D ǫ ∂t ∂ = 2E · D ∂t

Using this equation to replace the factor in the parentheses in the second term of Equation 3.9, we obtain:   ∂ E · (∇ × H) = H · − B − ∇ · (E × H) (3.11) ∂t

(3.3) Substituting this expression for the left side of Equation 3.7, we have

Summarizing:

−H·

∂ ∂ (E · D) = 2E · D ∂t ∂t

∂ ∂ B − ∇ · (E × H) = E · J + E · D (3.12) ∂t ∂t

(3.4) Next we invoke the identities of Equations 3.4 and 3.5 to replace the first and last terms: which is the expression we seek. It is worth noting 1 ∂ that the expressions on both sides of the equation (H · B) − ∇ · (E × H) − 2 ∂t have the same units, namely, those of power. 1 ∂ =E·J+ (E · D) (3.13) Using the same reasoning, we find: 2 ∂t Now we move the first term to the right-hand side and (3.5) integrate both sides over the volume V: Z where H is the magnetic field intensity (SI base units − ∇ · (E × H) dv V of A/m) and B is the magnetic flux density (SI base Z units of T). = E · J dv V Z Derivation of the theorem. We begin with the 1 ∂ + (E · D) dv differential form of Ampere’s law: 2 ∂t ZV 1 ∂ ∂ + (H · B) dv (3.14) ∇×H=J+ D (3.6) V 2 ∂t ∂t ∂ ∂ (H · B) = 2H · B ∂t ∂t

Taking the dot product with E on both sides: E · (∇ × H) = E · J + E ·

∂ D ∂t

The left side may be transformed into a surface integral using the divergence theorem: Z I (3.7) ∇ · (E × H) dv = (E × H) · ds (3.15) V

Let’s deal with the left side of this equation first. For this, we will employ a vector identity (Equation B.27, Appendix B.3). This identity states that for any two vector fields F and G,

S

where S is the closed surface that bounds V and ds is the outward-facing normal to this surface, as indicated in Figure 3.1. Finally, we exchange the order of time differentiation and volume integration in ∇ · (F × G) = G · (∇ × F) − F · (∇ × G) (3.8) the last two terms: I − (E × H) · ds Substituting E for F and H for G and rearranging S Z terms, we find = E · J dv V E · (∇ × H) = H · (∇ × E) − ∇ · (E × H) (3.9) Z 1 ∂ E · D dv + Next, we invoke the Maxwell-Faraday equation: 2 ∂t V Z 1 ∂ + H · B dv (3.16) ∂ 2 ∂t V ∇×E=− B (3.10) ∂t

3.1. POYNTING’S THEOREM

27

Equation 3.16 is Poynting’s theorem. Each of the four terms has the particular physical interpretation identified in Equation 3.1, as we will now demonstrate. Power dissipated by ohmic loss. The first term of the right side of Equation 3.16 is PΩ ,

Z

V

E · J dv

PM ,

PM

V

V

PΩ is due to the conversion of the electric field into conduction current, and subsequently into heat. This mechanism is commonly known as ohmic loss or joule heating. It is worth noting that this expression has the expected units. For example, in Equation 3.17, E (SI base units of V/m) times J (SI base units of A/m2 ) yields a quantity with units of W/m3 ; i.e., power per unit volume, which is power density. Then integration over volume yields units of W; hence, power. Energy storage in electric and magnetic fields. The second term on the right side of Equation 3.16 is: Z

V

E · D dv

(3.19)

Recall D = ǫE, where ǫ is the permittivity (SI base units of F/m) of the material. Thus, we may rewrite the previous equation as follows: Z 1 ∂ PE = E · ǫE dv 2 ∂t V Z 1 ∂ 2 = ǫ |E| dv 2 ∂t V  Z  ∂ 1 2 = ǫ |E| dv (3.20) ∂t 2 V The quantity in parentheses is We , the energy stored in the electric field within V.

1 ∂ 2 ∂t

Z

V

H · B dv

(3.21)

Recall B = µH, where µ is the permeability (SI base units of H/m) of the material. Thus, we may rewrite the previous equation as follows:

(3.17)

Equation 3.17 is Joule’s law. Joule’s law gives the power dissipated due to finite conductivity of material. The role of conductivity (σ, SI base units of S/m) can be made explicit using the relationship J = σE (Ohm’s law) in Equation 3.17, which yields: Z Z 2 PΩ = E · σE dv = σ |E| dv (3.18)

1 ∂ PE , 2 ∂t

The third term on the right side of Equation 3.16 is:

Z 1 ∂ H · µH dv = 2 ∂t V Z 1 ∂ 2 = µ |H| dv 2 ∂t V  Z  ∂ 1 2 = µ |H| dv ∂t 2 V

(3.22)

The quantity in parentheses is Wm , the energy stored in the magnetic field within V. Power flux through S. The left side of Equation 3.16 is I (3.23) Pnet,in , − (E × H) · ds S

The quantity E × H is the Poynting vector, which quantifies the spatial power density (SI base units of W/m2 ) of an electromagnetic wave and the direction in which it propagates. The reader has likely already encountered this concept. Regardless, we’ll confirm this interpretation of the quantity E × H in Section 3.2. For now, observe that integration of the Poynting vector over S as indicated in Equation 3.23 yields the total power flowing out of V through S. The negative sign in Equation 3.23 indicates that the combined quantity represents power flow in to V through S. Finally, note the use of a single quantity Pnet,in does not imply that power is entirely inward-directed or outward-directed. Rather, Pnet,in represents the net flux; i.e., the sum of the inwardand outward-flowing power. Summary. Equation 3.16 may now be stated concisely as follows: Pnet,in = PΩ +

∂ ∂ We + Wm ∂t ∂t

(3.24)

CHAPTER 3. WAVE PROPAGATION IN GENERAL MEDIA

28

Poynting’s theorem (Equation 3.24, with Equations 3.23, 3.17, 3.19, and 3.21) states that the net electromagnetic power flowing into a region of space may be either dissipated, or used to change the energy stored in electric and magnetic fields within that region. Since we derived this result directly from the general form of Maxwell’s equations, these three possibilities are all the possibilities allowed by classical physics, so this is a statement of conservation of power. Finally, note the theorem also works in reverse; i.e., the net electromagnetic power flowing out of a region of space must have originated from some active source (i.e., the reverse of power dissipation) or released from energy stored in electric or magnetic fields.

3.2 Poynting Vector [m0122]

In this section, we use Poynting’s theorem (Section 3.1) to confirm the interpretation of the Poynting vector S,E×H

(3.25)

as the spatial power density (SI base units of W/m2 ) and the direction of power flow. Figure 3.2 shows a uniform plane wave incident on a homogeneous and lossless region V defined by the closed right cylinder S. The axis of the cylinder is ˆ of the aligned along the direction of propagation k plane wave. The ends of the cylinder are planar and ˆ perpendicular to k.

Additional Reading: Now let us apply Poynting’s theorem: • “Poynting vector” on Wikipedia.

Pnet,in = PΩ +

∂ ∂ We + Wm ∂t ∂t

(3.26)

Since the region is lossless, PΩ = 0. Presuming no other electric or magnetic fields, We and Wm must also be zero. Consequently, Pnet,in = 0. However, this does not mean that there is no power flowing into V. Instead, Pnet,in = 0 merely indicates that the net power flowing into V is zero. Thus, we might instead express the result for the present scenario as follows: Pnet,in = Pin − Pout = 0

(3.27)

c C. Wang CC BY-SA 4.0

Figure 3.2: A uniform plane wave incident on a region bounded by the cylindrical surface S.

3.2. POYNTING VECTOR where Pin and Pout indicate power flow explicitly into and explicitly out of V as separate quantities. Proceeding, let’s ignore what we know about power flow in plane waves, and instead see where Poynting’s theorem takes us. Here, I Pnet,in , − (E × H) · ds = 0 (3.28) S

The surface S consists of three sides: The two flat ends, and the curved surface that connects them. Let us refer to these sides as S1 , S2 , and S3 , from left to right as shown in Figure 3.2. Then, Z − (E × H) · ds (3.29) S Z 1 − (E × H) · ds Z S2 − (E × H) · ds = 0 S3

On the curved surface S2 , ds is everywhere ˆ perpendicular to E × H (i.e., perpendicular to k). Therefore, the second integral is equal to zero. On the left end S1 , the outward-facing differential surface ˆ element ds = −kds. On the right end S3 , ˆ ds = +kds. We are left with: Z ˆ (E × H) · kds (3.30) S1 Z ˆ =0 − (E × H) · kds S3

ˆ is, by Compare this to Equation 3.27. Since k definition, the direction in which E × H points, the first integral must be Pin and the second integral must be Pout . Thus, Z ˆ Pin = (E × H) · kds (3.31) S1

and it follows that the magnitude and direction of E × H, which we recognize as the Poynting vector, are spatial power density and direction of power flow, respectively. The Poynting vector is named after English physicist J.H. Poynting, one of the co-discoverers of the concept. The fact that this vector points in the direction of power flow, and is therefore also a “pointing vector,” is simply a remarkable coincidence.

29 Additional Reading: • “Poynting vector” on Wikipedia.

CHAPTER 3. WAVE PROPAGATION IN GENERAL MEDIA

30

3.3 Wave Equations for Lossy Regions [m0128]

The wave equations for electromagnetic propagation in lossless and source-free media, in differential phasor form, are: e + ω 2 µǫE e =0 ∇2 E e + ω 2 µǫH e =0 ∇2 H

eimp represents impressed sources of current where J eind represents current which is induced by loss. and J An impressed current is one whose behavior is independent of other features, analogous to an independent current source in elementary circuit theory. In the absence of such sources, Equation 3.39 becomes: e = 0 + σE e J

(3.40)

e = σE e + jωǫE e ∇×H e = (σ + jωǫ) E

(3.41)

(3.32) Equation 3.38 may now be rewritten as: (3.33)

The constant ω 2 µǫ is labeled β 2 , and β turns out to be the phase propagation constant. Now, we wish to upgrade these equations to account for the possibility of loss. First, let’s be clear on what we mean by “loss.” Specifically, we mean the possibility of conversion of energy from the propagating wave into current, and subsequently to heat. This mechanism is described by Ohm’s law:

e = jωǫc E

(3.42) (3.43)

where we defined the new constant ǫc as follows: ǫc , ǫ − j

σ ω

(3.44)

This constant is known as complex permittivity. In the lossless case, ǫc = ǫ; i.e., the imaginary part of ǫc → 0 so there is no difference between the physical e = σE e J (3.34) permittivity ǫ and ǫc . The effect of the material’s loss is represented as a non-zero imaginary component of e is conduction current where σ is conductivity and J the permittivity. density (SI base units of A/m2 ). In the lossless case, σ is presumed to be zero (or at least negligible), so J is It is also common to express ǫc as follows: presumed to be zero. To obtain wave equations for ǫc = ǫ′ − jǫ′′ (3.45) media exhibiting significant loss, we cannot assume J = 0. where the real-valued constants ǫ′ and ǫ′′ are in this To obtain equations that account for the possibility of case: significant conduction current, hence loss, we return to the phasor forms of Maxwell’s equations: e = ρev ∇·E ǫ e = −jωµH e ∇×E e =0 ∇·H

ǫ′ = ǫ σ ǫ′′ = ω

(3.46) (3.47)

(3.35)

(3.36) This alternative notation is useful for three reasons. First, some authors use the symbol ǫ to refer to both (3.37) physical permittivity and complex permittivity. In this ′ ′′ e =J e + jωǫE e ∇×H (3.38) case, the “ǫ − jǫ ” notation is helpful in mitigating confusion. Second, it is often more convenient to specify ǫ′′ at a frequency than it is to specify σ, which where ρev is the charge density. If the region of interest is source-free, then ρev = 0. However, we may may also be a function of frequency. In fact, in some e since σ may be non-zero. To applications the loss of a material is most not similarly suppress J conveniently specified using the ratio ǫ′′ /ǫ′ (known as make progress, let us identify the possible e loss tangent, for reasons explained elsewhere). contributions to J as follows: Finally, it turns out that nonlinearity of permittivity e=J eimp + J eind J (3.39) can also be accommodated as an imaginary

3.3. WAVE EQUATIONS FOR LOSSY REGIONS component of the permittivity. The “ǫ′′ ” notation allows us to accommodate both effects – nonlinearity and conductivity – using common notation. In this section, however, we remain focused exclusively on conductivity. Complex permittivity ǫc (SI base units of F/m) describes the combined effects of permittivity and conductivity. Conductivity is represented as an imaginary-valued component of the permittivity. Returning to Equations 3.35–3.38, we obtain: e =0 ∇·E e = −jωµH e ∇×E

e =0 ∇·H e = jωǫc E e ∇×H

(3.48) (3.49) (3.50) (3.51)

31 phase of a wave progresses with increasing distance along the direction of propagation. Given the similarity of Equations 3.55 and 3.56 to Equations 3.32 and 3.33, respectively, the constant γ must play a similar role. So, we are motivated to find an expression for γ. At first glance, this is simple: p has two γ = γ 2 . However, recall that every number p √ square roots. When one declares β = β 2 = ω µǫ, there is no concern since β is, by definition, positive; therefore, one knows to take the positive-valued square root. In contrast, γ 2p is complex-valued, and so the two possible values of γ 2 are both potentially complex-valued. We are left without clear guidance on which of these values is appropriate and physically-relevant. So we proceed with caution. First, consider the special case that γ is purely imaginary; e.g., γ = jγ ′′ where γ ′′ is a real-valued 2 constant. In this case, γ 2 = − (γ ′′ ) and the wave equations become

These equations are identical to the corresponding equations for the lossless case, with the exception that ǫ has been replaced by ǫc . Similarly, we may replace the factor ω 2 µǫ in Equations 3.32 and 3.33, yielding: e + ω 2 µǫc E e =0 ∇2 E e + ω 2 µǫc H e =0 ∇2 H

(3.52) (3.53)

In the lossless case, ω 2 µǫc → ω 2 µǫ, which is β 2 as expected. For the general (i.e., possibly lossy) case, we shall make an analogous definition 2

2

γ , −ω µǫc

(3.54)

such that the wave equations may now be written as follows: e − γ2E e =0 (3.55) ∇2 E e − γ2H e =0 ∇2 H

(3.56)

e + (γ ′′ )2 E e =0 ∇2 E ′′ 2

(3.57)

e + (γ ) H e =0 ∇2 H

(3.58)

β , Im {γ}

(3.59)

Comparing to Equations 3.32 and 3.33, we see γ ′′ plays the exact same role as β; i.e., γ ′′ is the phase propagation constant for whatever wave we obtain as the solution to the above equations. Therefore, let us make that definition formally:

Be careful: Note that we are not claiming that γ ′′ in the possibly-lossy case is equal to ω 2 µǫ. Instead, we are asserting exactly the opposite; i.e., that there is a phase propagation constant in the general (possibly lossy) case, and we should find that this constant simplifies to ω 2 µǫ in the lossless case.

Note the minus sign in Equation 3.54, and the associated changes of signs in Equations 3.55 and 3.56. For the moment, this choice of sign may be viewed as arbitrary – we would have been equally justified in choosing the opposite sign for the definition of γ 2 . However, the choice we have made is customary, and yields some notational convenience that will become apparent later.

Now we make the following definition for the real component of γ:

In the lossless case, β is the phase propagation constant, which determines the rate at which the

Now, it is possible to determine γ explicitly in terms of ω and the constitutive properties of the material.

α , Re {γ}

(3.60)

γ = α + jβ

(3.61)

Such that where α and β are real-valued constants.

CHAPTER 3. WAVE PROPAGATION IN GENERAL MEDIA

32

“Wave Propagation on a TEM Transmission Line,”1 we found that the potential and current along a 2 γ 2 = (α + jβ) transverse electromagnetic (TEM) transmission line = α2 − β 2 + j2αβ (3.62) satisfy the same wave equations that we have developed in this section, having a complex-valued Expanding Equation 3.54 using Equations 3.45–3.47, propagation constant γ = α + jβ, and the same physical interpretation of β as the phase propagation we obtain:   constant. As is explained in another section, α too has σ γ 2 = −ω 2 µ ǫ − j the same interpretation in both applications – that is, ω as the attenuation constant. 2 = −ω µǫ + jωµσ (3.63) First, note:

The real and imaginary parts of Equations 3.62 and 3.63 must be equal. Enforcing this equality yields the following equations: α2 − β 2 = −ω 2 µǫ

(3.64)

2αβ = ωµσ

(3.65)

Additional Reading:

• “Electromagnetic Wave Equation” on Wikipedia.

Equations 3.64 and 3.65 are independent simultaneous equations that may be solved for α and β. Sparing the reader the remaining steps, which are purely mathematical, we find:  s 1/2  ′′ 2   µǫ′ ǫ  1+  − 1 α=ω   2 ǫ′  s 1/2  ′′ 2  µǫ′   1+ ǫ  β=ω + 1  2  ǫ′

(3.66)

(3.67)

Equations 3.66 and 3.67 can be verified by confirming that Equations 3.64 and 3.65 are satisfied. It is also useful to confirm that the expected results are obtained in the lossless case. In the lossless case, σ = 0, so ǫ′′ = 0. Subsequently, Equation 3.66 yields √ α = 0 and Equation 3.67 yields β = ω µǫ, as expected. The electromagnetic wave equations accounting for the possibility of lossy media are Equations 3.55 and 3.56 with γ = α + jβ, where α and β are the positive real-valued constants determined by Equations 3.66 and 3.67, respectively. We conclude this section by pointing out a very useful analogy to transmission line theory. In the section

1 This section may appear in a different volume, depending on the version of this book.

3.4. COMPLEX PERMITTIVITY

33

3.4 Complex Permittivity

where ǫc is a complex-valued constant that depends on frequency.

[m0134]

The notation “ǫc ” is used elsewhere in this book (e.g., Section 3.3) to represent a generalization of simple permittivity that accommodates loss associated with non-zero conductivity σ. In Section 3.3, we defined

The relationship between electric field intensity E (SI base units of V/m) and electric flux density D (SI base units of C/m2 ) is: D = ǫE

(3.68)

ǫc , ǫ′ − jǫ′′

(3.72)

where ǫ′ , ǫ (i.e., real-valued, simple permittivity) where ǫ is the permittivity (SI base units of F/m). In and ǫ′′ , σ/ω (i.e., the effect of non-zero simple media, ǫ is a real positive value which does not conductivity). Similarly, ǫc in Equation 3.71 can be depend on the time variation of E. That is, the expressed as response (D) to a change in E is observed ǫc , ǫ′ − jǫ′′ (3.73) instantaneously and without delay. but in this case In practical materials, however, the change in D in ǫ′ , a0 − ω 2 a2 + ... (3.74) response to a change in E may depend on the manner in which E changes. The response may not be and instantaneous, but rather might take some time to ǫ′′ , −ωa1 + ω 3 a3 − ... (3.75) fully manifest. This behavior can be modeled using the following generalization of Equation 3.68: When expressed as phasors, the temporal rela∂ ∂2 ∂3 tionship between electric flux density and elecD = a0 E+a1 E+a2 2 E+a3 3 E+... (3.69) ∂t ∂t ∂t tric field intensity can be expressed as a complexvalued permittivity. where a , a , a , and so on are real-valued constants, 0

1

2

and the number of terms is infinite. In practical materials, the importance of the terms tends to diminish with increasing order. Thus, it is common that only the first few terms are significant. In many applications involving common materials, only the first term is significant; i.e., a0 ≈ ǫ and an ≈ 0 for n ≥ 1. As we shall see in a moment, this distinction commonly depends on frequency.

Thus, we now see two possible applications for the concept of complex permittivity: Modeling the delayed response of D to changing E, as described above; and modeling loss associated with non-zero conductivity. In practical work, it may not always be clear precisely what combination of effects the complex permittivity ǫc = ǫ′ − jǫ′′ is taking into account. For example, if ǫc is obtained by In the phasor domain, differentiation with respect to measurement, both delayed response and conduction time becomes multiplication by jω. Thus, loss may be represented. Therefore, it is not Equation 3.69 becomes reasonable to assume a value of ǫ′′ obtained by e = a0 E e + a1 (jω) E e + a2 (jω)2 E e + a3 (jω)3 E e + ... measurement represents only conduction loss (i.e., is D equal to σ/ω); in fact, the measurement also may e + jωa1 E e − ω 2 a2 E e − jω 3 a3 E e + ... = a0 E include a significant frequency-dependent  2 3 e = a0 + jωa1 − ω a2 − jω a3 + ... E (3.70) contribution associated with the delayed response behavior identified in this section. Note that the factor in parentheses is a complex-valued number which depends on frequency An example of the complex permittivity of a typical dielectric material is shown in Figure 3.3. Note the ω and materials parameters a0 , a1 , ... We may frequency dependence is quite simple and summarize this as follows: slowly-varying at frequencies below 1 GHz or so, but e = ǫc E e D (3.71) becomes relatively complex at higher frequencies.

CHAPTER 3. WAVE PROPAGATION IN GENERAL MEDIA

34

3.5 Loss Tangent [m0132]

In Section 3.3, we found that the effect of loss due to non-zero conductivity σ could be concisely quantified using the ratio ǫ′′ σ = (3.76) ǫ′ ωǫ where ǫ′ and ǫ′′ are the real and imaginary components of the complex permittivity ǫc , and ǫ′ , ǫ. In this section, we explore this relationship in greater detail. c K.A. Mauritz (modified)

Recall Ampere’s law in differential (but otherwise Figure 3.3: The relative contributions of the real and general) form: imaginary components of permittivity for a typical dielectric material (in this case, a polymer). ∂ ∇×H=J+ D (3.77) ∂t Additional Reading: • “Permittivity” on Wikipedia.

The first term on the right is conduction current, whereas the second term on the right is displacement current. In the phasor domain, differentiation with respect to time (∂/∂t) becomes multiplication by jω. Therefore, the phasor form of Equation 3.77 is e =J e + jω D e ∇×H

(3.78)

e = σE e + jωǫE e ∇×H

(3.79)

e = σE e and that D e = ǫE. e Thus, Also recall that J Interestingly, the total current is the sum of a real-valued conduction current and an imaginary-valued displacement current. This is shown graphically in Figure 3.4.

Note that the angle δ indicated in Figure 3.4 is given by σ tan δ , (3.80) ωǫ The quantity tan δ is referred to as the loss tangent. Note that loss tangent is zero for a lossless (σ ≡ 0) material, and increases with increasing loss. Thus, loss tangent provides an alternative way to quantify the effect of loss on the electromagnetic field within a material. Loss tangent presuming only ohmic (conduction) loss is given by Equation 3.80.

3.5. LOSS TANGENT

35

Comparing Equation 3.80 to Equation 3.76, we see loss tangent can equivalently be calculated as tan δ =

ǫ′′ ǫ

(3.81)

and subsequently interpreted as shown in Figure 3.5. The discussion in this section has assumed that ǫc is complex-valued solely due to ohmic loss. However, it is explained in Section 3.4 that permittivity may also be complex-valued as a way to model delay in the response of D to changing E. Does the concept of loss tangent apply also in this case? Since the math does not distinguish between permittivity which is complex due to loss and permittivity which is complex due to delay, subsequent mathematically-derived results apply in either case. On the other hand, there may be potentially significant differences in the physical manifestation of these effects. For example, a material having large loss tangent due to ohmic loss might become hot when a large electric field is applied, whereas a material having large loss tangent due to delayed response might not. Summarizing: The expression for loss tangent given by Equation 3.81 and Figure 3.5 does not distinguish between ohmic loss and delayed response.

c C. Wang CC BY-SA 4.0

Figure 3.4: In the phasor domain, the total current is the sum of a real-valued conduction current and an imaginary-valued displacement current.

c C. Wang CC BY-SA 4.0

Figure 3.5: Loss tangent defined in terms of the real and imaginary components of the complex permittivity ǫc . Additional Reading: • “Dielectric loss” on Wikipedia.

CHAPTER 3. WAVE PROPAGATION IN GENERAL MEDIA

36

3.6 Plane Waves in Lossy Regions

the voltage and current in lossy transmission lines, as described in the section “Wave Equation for a TEM Transmission Line.”3

[m0130]

The electromagnetic wave equations for source-free regions consisting of possibly-lossy material are (see Section 3.3):

where

e − γ2E e =0 ∇2 E e − γ2H e =0 ∇2 H γ 2 , −ω 2 µǫc

ˆ -polarized plane Let’s consider the special case of an x wave propagating in the +ˆ z direction: + −γz e =x ˆ Ex0 E e

(3.88)

γ = α + jβ

(3.89)

(3.82) We established in Section 3.3 that γ may be written explicitly in terms of its real and imaginary (3.83) components as follows: (3.84)

We now turn our attention to the question, what are the characteristics of waves that propagate in these conditions? As in the lossless case, these equations permit waves having a variety of geometries including plane waves, cylindrical waves, and spherical waves. In this section, we will consider the important special case of uniform plane waves.

where α and β are positive real-valued constants depending on frequency (ω) and constitutive properties of the medium; i.e., permittivity, permeability, and conductivity. Thus:

To obtain the general expression for the uniform plane wave solution, we follow precisely the same procedure described in the section “Uniform Plane Waves: Derivation.”2 Although that section presumed lossless media, the only difference in the present situation is that the real-valued constant +β 2 is replaced with the complex-valued constant −γ 2 . Thus, we obtain the desired solution through a simple modification of the solution for the lossless case. For a wave exhibiting uniform magnitude and phase in planes of constant z, we find that the electric field is:

Observe that the variation of phase with distance is determined by β through the factor e−jβz ; thus, β is the phase propagation constant and plays precisely the same role as in the lossless case. Observe also that the variation in magnitude is determined by α through the real-valued factor e−αz . Specifically, magnitude is reduced inverse-exponentially with increasing distance along the direction of propagation. Thus, α is the attenuation constant and plays precisely the same role as the attenuation constant for a lossy transmission line.

e =x ex + y ey ˆE ˆE E

(3.85)

ex = E + e−γz + E − e+γz E x0 x0 ey = E + e−γz + E − e+γz E y0 y0

The presence of loss in material gives rise to a real-valued factor e−αz which describes the attenuation of the wave with propagation (in this case, along z) in the material.

(3.86)

where

+ −(α+jβ)z e =x ˆ Ex0 E e

+ −αz −jβz ˆ Ex0 =x e e

(3.90)

(3.87) We may continue to exploit the similarity of the potentially-lossy and lossless plane wave results to + − where the complex-valued coefficients Ex0 , Ex0 , quickly ascertain the characteristics of the magnetic + − Ey0 , and Ey0 are determined by boundary conditions field. In particular, the plane wave relationships apply (possibly including sources) outside the region of exactly as they do in the lossless case. These interest. This result can be confirmed by verifying relationships are: that Equations 3.86 and 3.87 each satisfy Equation 3.82. Also, it may be helpful to note that ˆ×E e = 1k e H (3.91) these expressions are identical to those obtained for η 2 Depending on the version of this book, this section may appear in a different volume.

3 Depending on the version of this book, this section may appear in a different volume.

3.7. WAVE POWER IN A LOSSY MEDIUM ˆ×H e = −η k e E

37 (3.92)

ˆ is the direction of propagation and η is the where k p wave impedance. In the lossless case, η = µ/ǫ; however, in the possibly-lossy case we must replace ǫ = ǫ′ with ǫc = ǫ′ − jǫ′′ . Thus: r r µ µ η → ηc = = ′ ǫc ǫ − jǫ′′ r s µ 1 (3.93) = ′ ǫ 1 − j (ǫ′′ /ǫ′ ) Thus: ηc =

r

 −1/2 ǫ′′ µ · 1−j ′ ǫ′ ǫ

(3.94)

Remarkably, we find that p the wave impedance for a lossy material is equal to µ/ǫ – the wave impedance we would calculate if we neglected loss (i.e., assumed σ = 0) – times a correction factor that accounts for the loss. This correction factor is complex-valued; therefore, E and H are not in phase when propagating through lossy material. We now see that in the phasor domain: ˆ×E e e = 1k H ηc

ˆ×H e = −ηc k e E

(3.95) (3.96)

The plane wave relationships in media which are possibly lossy are given by Equations 3.95 and 3.96, with the complex-valued wave impedance given by Equation 3.94.

3.7 Wave Power in a Lossy Medium [m0133]

In this section, we consider the power associated with waves propagating in materials which are potentially lossy; i.e., having conductivity σ significantly greater than zero. This topic has previously been considered in the section “Wave Power in a Lossless Medium” for the case in loss is not significant.4 A review of that section may be useful before reading this section. Recall that the Poynting vector S,E×H

(3.97)

indicates the power density (i.e., W/m2 ) of a wave and the direction of power flow. This is “instantaneous” power, applicable to waves regardless of the way they vary with time. Often we are interested specifically in waves which vary sinusoidally, and which subsequently may be represented as phasors. In this case, the time-average Poynting vector is 1 n e e ∗o Save , Re E ×H (3.98) 2

Further, we have already used this expression to find that the time-average power density for a sinusoidally-varying uniform plane wave in a lossless medium is simply Save =

|E0 | 2η

2

(lossless case)

(3.99)

where |E0 | is the peak (as opposed to RMS) magnitude of the electric field intensity phasor, and η is the wave impedance. Let us now use Equation 3.98 to determine the expression corresponding to Equation 3.99 in the case of possibly-lossy media. We may express the electric and magnetic field intensities of a uniform plane wave as e =x ˆ E0 e−αz e−jβz E (3.100) and

E0 e =y ˆ e−αz e−jβz H ηc

(3.101)

4 Depending on the version of this book, this section may appear in another volume.

CHAPTER 3. WAVE PROPAGATION IN GENERAL MEDIA

38

where α and β are the attenuation constant and phase propagation constant, respectively, and ηc is the complex-valued wave impedance. As written, these expressions describe a wave which is +ˆ x-polarized and propagates in the +ˆ z direction. We make these choices for convenience only – as long as the medium is homogeneous and isotropic, we expect our findings to apply regardless of polarization and direction of propagation. Applying Equation 3.98: 1 n e e ∗o Save = Re E ×H 2 ( ) 2 |E0 | −2αz 1 ˆ Re e = z 2 ηc∗   2 |E0 | 1 ˆ =z Re e−2αz (3.102) 2 ηc∗ Because ηc is complex-valued when the material is lossy, we must proceed with caution. First, let us write ηc explicitly in terms of its magnitude |ηc | and phase ψη : η , |η| ejψη (3.103) Then:

ηc∗ = |ηc | e−jψη

n

−1 (ηc∗ )

Re (ηc∗ )

−1

o

= |ηc | = |ηc |

(3.104)

−1 +jψη

e

(3.105)

−1

cos ψη

(3.106)

Then Equation 3.102 may be written: 2

ˆ Save = z

|E0 | −2αz e cos ψη 2 |ηc |

(3.107)

The time-average power density of the plane wave described by Equation 3.100 in a possiblylossy material is given by Equation 3.107. As a check, note that this expression gives the expected result for lossless media; i.e., α = 0, |ηc | = η, and ψη = 0. We now see that the effect of loss is that power density is now proportional to 2 (e−αz ) , so, as expected, power density is proportional to the square of either |E| or |H|. The result further indicates a one-time scaling of the power density by a factor of |η| cos ψη < 1 |ηc |

relative to a medium without loss.

(3.108)

The reduction in power density due to nonzero conductivity is proportional to a distancedependent factor e−2αz and an additional factor that depends on the magnitude and phase of ηc .

3.8. DECIBEL SCALE FOR POWER RATIO

3.8 Decibel Scale for Power Ratio [m0154]

In many disciplines within electrical engineering, it is common to evaluate the ratios of powers and power densities that differ by many orders of magnitude. These ratios could be expressed in scientific notation, but it is more common to use the logarithmic decibel (dB) scale in such applications. In the conventional (linear) scale, the ratio of power P1 to power P0 is simply G=

P1 (linear units) P0

39 Example 3.1. Power loss from a long cable. A 2 W signal is injected into a long cable. The power arriving at the other end of the cable is 10 µW. What is the power loss in dB? In linear units: G=

10 µW = 5 × 10−6 (linear units) 2W

In dB:  G = 10 log10 5 × 10−6 ∼ = −53.0 dB ∼ +53.0 dB L = −G =

(3.109)

The decibel scale is used in precisely the same way to relate ratios of spatial power densities for waves. For example, the loss incurred when the spatial power density is reduced from S0 (SI base units of W/m2 ) to In the decibel scale, the ratio of power P1 to power P0 S is 1 is S0 L = 10 log10 (3.111) (dB) P1 S 1 G , 10 log10 (3.110) (dB) P0 This works because the common units of m−2 in the numerator and denominator cancel, leaving a power where “dB” denotes a unitless quantity which is ratio. expressed in the decibel scale. Note that G < 0 dB (i.e., is “negative in dB”) if P1 < P0 and G > 0 dB if A common point of confusion is the proper use of the P1 > P0 . decibel scale to represent voltage or current ratios. To avoid confusion, simply refer to the definition The power gain P1 /P0 in dB is given by Equaexpressed in Equation 3.110. For example, let’s say tion 3.110. P1 = V12 /R1 where V1 is potential and R1 is the impedance across which V1 is defined. Similarly, let Alternatively, one might choose to interpret a power us define P0 = V02 /R0 where V0 is potential and R0 ratio as a loss L with L , 1/G in linear units, which is the impedance across which V0 is defined. is L = −G when expressed in dB. Most often, but not Applying Equation 3.110: always, engineers interpret a power ratio as “gain” if P1 the output power is expected to be greater than input (dB) G , 10 log10 P0 power (e.g., as expected for an amplifier) and as V 2 /R1 “loss” if output power is expected to be less than = 10 log10 12 (dB) (3.112) input power (e.g., as expected for a lossy transmission V0 /R0 line). Now, if R1 = R0 , then Power loss L is the reciprocal of power gain G. V2 G = 10 log10 12 (dB) Therefore, L = −G when these quantities are V0 expressed in dB.  2 V1 = 10 log10 (dB) V0 V1 (3.113) = 20 log10 (dB) V0 Here, “G” might be interpreted as “power gain.” Note that G < 1 if P1 < P0 and G > 1 if P1 > P0 .

CHAPTER 3. WAVE PROPAGATION IN GENERAL MEDIA

40

3.9 Attenuation Rate

However, note that this is not true if R1 6= R0 . A power ratio in dB is equal to 20 log10 of the voltage ratio only if the associated impedances are equal. Adding to the potential for confusion on this point is the concept of voltage gain Gv : Gv , 20 log10

V1 (dB) V0

Attenuation rate is a convenient way to quantify loss in general media, including transmission lines, using the decibel scale.

Consider a transmission line carrying a wave in the +z direction. Let P0 be the power at z = 0. Let P1 be (3.114) the power at z = l. Then the power at z = 0 relative to the power at z = l is:

which applies regardless of the associated impedances. Note that Gv = G only if the associated impedances are equal, and that these ratios are different otherwise. Be careful! The decibel scale simplifies common calculations. Here’s an example. Let’s say a signal having power P0 is injected into a transmission line having loss L. Then the output power P1 = P0 /L in linear units. However, in dB, we find: P0 L = 10 log10 P0 − 10 log10 L

10 log10 P1 = 10 log10

Division has been transformed into subtraction; i.e., P1 = P0 − L (dB)

[m0155]

(3.115)

This form facilitates easier calculation and visualization, and so is typically preferred. Finally, note that the units of P1 and P0 in Equation 3.115 are not dB per se, but rather dB with respect to the original power units. For example, if P1 is in mW, then taking 10 log10 of this quantity results in a quantity having units of dB relative to 1 mW. A power expressed in dB relative to 1 mW is said to have units of “dBm.” For example, “0 dBm” means 0 dB relative to 1 mW, which is simply 1 mW. Similarly +10 dBm is 10 mW, −10 dBm is 0.1 mW, and so on.

e−2α·0 P0 = −2α·l = e2αl (linear units) P1 e

(3.116)

where α is the attenuation constant; that is, the real part of the propagation constant γ = α + jβ. Expressed in this manner, the power ratio is a loss; that is, a number greater than 1 represents attenuation. In the decibel scale, the loss is 10 log10

P0 = 10 log10 e2αl P1 = 20αl log10 e ∼ = 8.69αl dB

(3.117)

Attenuation rate is defined as this quantity per unit length. Dividing by l, we obtain: attenuation rate ∼ = 8.69α

(3.118)

This has units of dB/length, where the units of length are the same length units in which α is expressed. For example, if α is expressed in units of m−1 , then attenuation rate has units of dB/m. Attenuation rate ∼ = 8.69α is the loss in dB, per unit length. The utility of the attenuation rate concept is that it allows us to quickly calculate loss for any distance of wave travel: This loss is simply attenuation rate (dB/m) times length (m), which yields loss in dB. Example 3.2. Attenuation rate in a long cable.

Additional Reading: • “Decibel” on Wikipedia. • “Scientific notation” on Wikipedia.

A particular coaxial cable has an attenuation constant α ∼ = 8.5 × 10−3 m−1 . What is the attenuation rate and the loss in dB for 100 m of this cable?

3.10. POOR CONDUCTORS The attenuation rate is ∼ = 8.69α ∼ = 0.0738 dB/m

41

3.10 Poor Conductors [m0156]

The loss in 100 m of this cable is ∼ = (0.0738 dB/m) (100 m) ∼ = 7.4 dB Note that it would be entirely appropriate, and equivalent, to state that the attenuation rate for this cable is 7.4 dB/(100 m). The concept of attenuation rate is used in precisely the same way to relate ratios of spatial power densities for unguided waves. This works because spatial power density has SI base units of W/m2 , so the common units of m−2 in the numerator and denominator cancel in the power density ratio, leaving a simple power ratio.

A poor conductor is a material for which conductivity is low, yet sufficient to exhibit significant loss. To be clear, the loss we refer to here is the conversion of the electric field to current through Ohm’s law. The threshold of significance depends on the application. For example, the dielectric spacer separating the conductors in a coaxial cable might be treated as lossless (σ = 0) for short lengths at low frequencies; whereas the loss of the cable for long lengths and higher frequencies is typically significant, and must be taken into account. In the latter case, the material is said to be a poor conductor because the loss is significant yet the material can still be treated in most other respects as an ideal dielectric. A quantitative but approximate criterion for identification of a poor conductor can be obtained from the concept of complex permittivity ǫc , which has the form: ǫc = ǫ′ − jǫ′′ (3.119) Recall that ǫ′′ quantifies loss, whereas ǫ′ exists independently of loss. In fact, ǫc = ǫ′ = ǫ for a perfectly lossless material. Therefore, we may quantify the lossiness of a material using the ratio ǫ′′ /ǫ′ , which is sometimes referred to as loss tangent (see Section 3.5). Using this quantity, we define a poor conductor as a material for which ǫ′′ is very small relative to ǫ′ . Thus, ǫ′′ ≪ 1 (poor conductor) ǫ′

(3.120)

A poor conductor is a material having loss tangent much less than 1, such that it behaves in most respects as an ideal dielectric except that ohmic loss may not be negligible. An example of a poor conductor commonly encountered in electrical engineering includes the popular printed circuit board substrate material FR4 (fiberglass epoxy), which has ǫ′′ /ǫ′ ∼ 0.008 over the frequency range it is most commonly used. Another example, already mentioned, is the dielectric spacer

CHAPTER 3. WAVE PROPAGATION IN GENERAL MEDIA

42

material (for example, polyethylene) typically used in coaxial cables. The loss of these materials may or may not be significant, depending on the particulars of the application. The imprecise definition of Equation 3.120 is sufficient to derive some characteristics exhibited by all poor conductors. To do so, first recall that the propagation constant γ is given in general as follows:

At this point, we are able to identify an expression for the phase propagation constant: p β , Im {γ} ≈ ω µǫ′ (poor conductor) (3.128) Remarkably, we find that β for a poor conductor is approximately equal to β for an ideal dielectric. For the attenuation constant, we find p

ǫ′′ (poor conductor) 2ǫ′ (3.129) Therefore: p γ = −ω 2 µǫc (3.122) Alternatively, this expression may be written in the following form: In general a number has two square roots, so some 1 ǫ′′ caution is required here. In this case, we may proceed α ≈ β ′ (poor conductor) (3.130) as follows: 2 ǫ √ p Presuming that ǫc is determined entirely by ohmic γ = jω µ ǫ′ − jǫ′′ loss, then r p ǫ′′ σ ǫ′′ ′ = (3.131) = jω µǫ 1 − j ′ (3.123) ′ ǫ ωǫ ǫ Under this condition, Equation 3.129 may be The requirement that ǫ′′ /ǫ′ ≪ 1 for a poor conductor rewritten: allows this expression to be “linearized.” For this, we p σ α ≈ ω µǫ′ (poor conductor) (3.132) invoke the binomial series representation: 2ωǫ ′ n(n − 1) 2 n x + ... (3.124) Since ǫ = ǫ under these assumptions, the expression (1 + x) = 1 + nx + 2! simplifies to r where x and n are, for our purposes, any constants; 1 σ µ = ση (poor conductor) (3.133) α≈ and “...” indicates the remaining terms in this infinite 2 ǫ 2 series, with each term containing the factor xn with p where η , µ/ǫ′ is the wave impedance presuming n > 2. If x ≪ 1, then all terms containing xn with n ≥ 2 will be very small relative to the first two terms lossless material. This result is remarkable for two reasons: First, factors of ω have been eliminated, so of the series. Thus, there is no dependence on frequency separate from n (1 + x) ≈ 1 + nx for x ≪ 1 (3.125) the frequency dependence of the constitutive parameters σ, µ, and ǫ. These parameters vary slowly Applying this to the present problem: with frequency, so the value of α for a poor conductor also varies slowly with frequency. Second, we see α   1/2 ǫ′′ ǫ′′ 1−j ′ ≈1−j ′ (3.126) is proportional to σ and η. This makes it quite easy to ǫ 2ǫ anticipate how the attenuation constant is affected by changes in conductivity and wave impedance in poor where we have used n = 1/2 and x = −jǫ′′ /ǫ′ . conductors. Applying this approximation to Equation 3.123, we obtain: Finally, what is the wave impedance in a poor   ′′ conductor? In contrast to η, ηc is potentially p ǫ γ ≈ jω µǫ′ 1 − j ′ complex-valued and may depend on σ. First, recall: 2ǫ r  −1/2 ′′ p p ǫ µ ǫ′′ ≈ jω µǫ′ + ω µǫ′ ′ (3.127) ηc = · 1−j ′ (3.134) 2ǫ ǫ′ ǫ γ 2 = −ω 2 µǫc

(3.121)

α , Re {γ} ≈ ω

µǫ′

3.11. GOOD CONDUCTORS

43

Applying the same approximation applied to γ earlier, this may be written r   µ ǫ′′ ηc ≈ · 1−j ′ (poor conductor) (3.135) ǫ′ 2ǫ

3.11 Good Conductors

Additional Reading:

Conductors are materials which are intended to efficiently sustain current, which requires high conductivity σ. In contrast, non-conductors are materials which are intended to efficiently sustain the electric field, which requires low σ. “Loss” for a non-conductor (see in particular “poor conductors,” Section 3.10) means the conversion of energy in the electric field into current. In contrast, “loss” for a conductor refers to energy already associated with current, which is subsequently dissipated in resistance. Summarizing: A good (“low-loss”) conductor is a material with high conductivity, such that power dissipated in the resistance of the material is low.

[m0157]

A good conductor is a material which behaves in most respects as a perfect conductor, yet exhibits We see that for a poor conductor, Re{ηc } ≈ η and significant loss. Now, we have to be very careful: The that Im{ηc } ≪ Re {ηc }. The usual approximation in term “loss” applied to the concept of a “conductor” this case is simply means something quite different from the term “loss” ηc ≈ η (poor conductor) (3.136) applied to other types of materials. Let us take a moment to disambiguate this term.

• “Binomial series” on Wikipedia.

A quantitative criterion for a good conductor can be obtained from the concept of complex permittivity ǫc , which has the form: ǫc = ǫ′ − jǫ′′

(3.137)

Recall that ǫ′′ is proportional to conductivity (σ) and so ǫ′′ is very large for a good conductor. Therefore, we may identify a good conductor using the ratio ǫ′′ /ǫ′ , which is sometimes referred to as “loss tangent” (see Section 3.5). Using this quantity we define a good conductor as a material for which: ǫ′′ ≫ 1 (good conductor) ǫ′

(3.138)

This condition is met for most materials classified as “metals,” and especially for metals exhibiting very high conductivity such as gold, copper, and aluminum. A good conductor is a material having loss tangent much greater than 1.

CHAPTER 3. WAVE PROPAGATION IN GENERAL MEDIA

44

The imprecise definition of Equation 3.138 is sufficient to derive some characteristics that are common to materials over a wide range of conductivity. To derive these characteristics, first recall that the propagation constant γ is given in general as follows: γ 2 = −ω 2 µǫc Therefore: γ=

p

for delayed polarization response (as described in Section 3.4), then σ ǫ′′ = (3.146) ω

Under this condition, Equation 3.144 may be rewritten: r r (3.139) µσ ωµσ = (good conductor) (3.147) α≈ω 2ω 2

−ω 2 µǫc

(3.140) Since ω = 2πf , another possible form for this expression is In general, a number has two square roots, so some caution is required here. In this case, we may proceed p α ≈ πf µσ (good conductor) (3.148) as follows: √ p γ = jω µ ǫ′ − jǫ′′ r p ǫ′′ ′ = jω µǫ 1 − j ′ ǫ

Since ǫ′′ /ǫ′ ≫ 1 for a good conductor, r p ǫ′′ ′ γ ≈ jω µǫ −j ′ ǫ p p ′′ −j ≈ jω µǫ

The conductivity of most materials changes very slowly with frequency, so this expression indicates (3.141) that α (and β) increases approximately in proportion to the square root of frequency for good conductors. This is commonly observed in electrical engineering applications. For example, the attenuation rate√of transmission lines increases approximately as f . This is so because the principal contribution to the attenuation is resistance in the conductors comprising (3.142) the line.

To proceed, we must determine the principal √value of √ √ The attenuation rate for signals conveyed by −j. The answer is that −j = (1 − j) / 2.5 transmission lines is approximately proportional Continuing: to the square root of frequency. r r µǫ′′ µǫ′′ γ ≈ jω +ω (3.143) 2 2 Let us now consider the wave impedance ηc in a good conductor. Recall: We are now able to identify expressions for the r  −1/2 µ ǫ′′ attenuation and phase propagation constants: · 1−j ′ (3.149) ηc = ǫ′ ǫ r µǫ′′ Applying the same approximation applied to γ earlier (good conductor) α , Re {γ} ≈ ω 2 in this section, the previous expression may be written (3.144) r  −1/2 β , Im {γ} ≈ α (good conductor) (3.145) µ ǫ′′ ηc ≈ · −j ′ (good conductor) ǫ′ ǫ Remarkably, we find that α ≈ β for a good conductor, r µ 1 and neither α nor β depend on ǫ′ . ≈ ·√ (3.150) ǫ′′ −j In the special case that ǫc is determined entirely by √ √ conductivity loss (i.e., σ > 0) and is not accounting We’ve already established that −j = (1 − j) / 2. Applying that result here: 5 You can confirm this simply by squaring this result. The easiest way to derive this result is to work in polar form, in which −j is 1 at √ r µ 2 an angle of −π/2, and the square root operation consists of taking ηc ≈ · (3.151) the square root of the magnitude and dividing the phase by 2. ǫ′′ 1 − j

3.11. GOOD CONDUCTORS Now multiplying numerator and denominator by 1 + j, we obtain r µ ηc ≈ · (1 + j) (3.152) 2ǫ′′

45 this distance is the skin depth of the material. The concept of skin depth is presented in Section 3.12.

The dependence of β on conductivity leads to a particularly surprising result for the phase velocity of the beleaguered waves that do manage to propagate In the special case that ǫc is determined entirely by within a good conductor. Recall that for both lossless conductivity loss and is not accounting for delayed and low-loss (“poor conductor”) materials, the phase √ polarization response, then ǫ′′ = σ/ω, and we find: velocity vp is either exactly or approximately c/ ǫr , r where ǫr , ǫ′ /ǫ0 , resulting in typical phase velocities µω (3.153) within one order of magnitude of c. For a good ηc ≈ · (1 + j) 2σ conductor, we find instead: ω ω There are at least two other ways in which this (good conductor) (3.156) vp = ≈ √ β πf µσ expression is commonly written. First, we can use ω = 2πf to obtain: and since ω = 2πf : s r 4πf πf µ vp ≈ (good conductor) (3.157) ηc ≈ · (1 + j) (3.154) µσ σ √ Note that the phase velocity in a good conductor Second, we can use the fact that α ≈ πf µσ for increases with frequency and decreases with good conductors to obtain: conductivity. In contrast to poor conductors and α ηc ≈ · (1 + j) (3.155) non-conductors, the phase velocity in good σ conductors is usually a tiny fraction of c. For example, for a non-magnetic (µ ≈ µ0 ) good In any event, we see that the magnitude of the wave impedance bears little resemblance to the wave conductor with typical σ ∼ 106 S/m, we find impedance for a poor conductor. In particular, there is vp ∼ 100 km/s at 1 GHz – just ∼ 0.03% of the speed no dependence on the physical permittivity ǫ′ = ǫ, as of light in free space. we saw also for α and β. In this sense, the concept of This result also tells us something profound about the permittivity does not apply to good conductors, and nature of signals that are conveyed by transmission especially so for perfect conductors. lines. Regardless of whether we analyze such signals as voltage and current waves associated with the Note also that ψη , the phase of ηc , is always ≈ π/4 conductors or in terms of guided waves between the for a good conductor, in contrast to ≈ 0 for a poor conductors, we find that the phase velocity is within conductor. This has two implications that are useful an order of magnitude or so of c. Thus, the to know. First, since ηc is the ratio of the magnitude information conveyed by signals propagating along of the electric field intensity to the magnitude of the transmission lines travels primarily within the space magnetic field intensity, the phase of the magnetic field will be shifted by ≈ π/4 relative to the phase of between the conductors, and not within the conductors. Information cannot travel primarily in the the electric field in a good conductor. Second, recall conductors, as this would then result in apparent from Section 3.7 that the power density for a wave is proportional to cos ψη . Therefore, the extent to which phase velocity which is orders of magnitude less than c, as noted previously. Remarkably, classical a good conductor is able to “extinguish” a wave transmission line theory employing the R′ , G′ , C ′ , L′ propagating through it is determined entirely by α, −αl equivalent circuit model6 gets this right, even though and specifically is proportional to e where l is distance traveled through the material. In other words, that approach does not explicitly consider the only a perfect conductor (σ → ∞) is able to possibility of guided waves traveling between the conductors. completely suppress wave propagation, whereas 6 Depending on the version of this book, this topic may appear in waves are always able to penetrate some distance into any conductor which is merely “good.” A measure of another volume.

CHAPTER 3. WAVE PROPAGATION IN GENERAL MEDIA

46

3.12 Skin Depth

Example 3.3. Skin depth of aluminum.

[m0158]

The electric and magnetic fields of a wave are diminished as the wave propagates through lossy media. The magnitude of these fields is proportional to e−αl where α , Re {γ} is the attenuation constant (SI base units of m−1 ), γ is the propagation constant, and l is the distance traveled. Although the rate at which magnitude is reduced is completely described by α, particular values of α typically do not necessarily provide an intuitive sense of this rate. An alternative way to characterize attenuation is in terms of skin depth δs , which is defined as the distance at which the magnitude of the electric and magnetic fields is reduced by a factor of 1/e. In other words: e−αδs = e−1 ∼ (3.158) = 0.368 Skin depth δs is the distance over which the magnitude of the electric or magnetic field is reduced by a factor of 1/e ∼ = 0.368. Since power is proportional to the square of the field magnitude, δs may also be interpreted as the distance at which the power in the wave is reduced by a factor of (1/e)2 ∼ = 0.135. In yet other words: δs is the distance at which ∼ = 86.5% of the power in the wave is lost. This definition for skin depth makes δs easy to compute: From Equation 3.158, it is simply δs =

1 α

(3.159)

The concept of skin depth is most commonly applied to good √ conductors. For a good conductor, α ≈ πf µσ (Section 3.11), so δs ≈ √

1 (good conductors) πf µσ

(3.160)

Aluminum, a good conductor, exhibits σ ≈ 3.7 × 107 S/m and µ ≈ µ0 over a broad range of radio frequencies. Using Equation 3.160, we find δs ∼ 26 µm at 10 MHz. Aluminum sheet which is 1/16-in (∼ = 1.59 mm) thick can also be said to have a thickness of ∼ 61δs at 10 MHz. The reduction in the power density of an electromagnetic wave after traveling through this sheet will be  2  2 ∼ e−α(61δs ) = e−α(61/α) = e−122

which is effectively zero from a practical engineering perspective. Therefore, 1/16-in aluminum sheet provides excellent shielding from electromagnetic waves at 10 MHz.

At 1 kHz, the situation is significantly different. At this frequency, δs ∼ 2.6 mm, so 1/16-in aluminum is only ∼ 0.6δs thick. In this case the power density is reduced by only  2  2 ∼ e−α(0.6δs ) = e−α(0.6/α) = e−1.2 ≈ 0.3 This is a reduction of only ∼ 70% in power density. Therefore, 1/16-in aluminum sheet provides very little shielding at 1 kHz.

[m0183]

3.12. SKIN DEPTH

Image Credits c Sevenchw (C. Wang), Fig. 3.1: https://commons.wikimedia.org/wiki/File:Poynting%E2%80%99s theorem illustration.svg, CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 3.2: https://commons.wikimedia.org/wiki/File:Uniform plane wave incident cylindrical surface.svg, CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c K.A. Mauritz, https://commons.wikimedia.org/wiki/File:Dielectric responses.svg, Fig. 3.3: Used with permission (see URL) and modified by author. c Sevenchw (C. Wang), Fig. 3.4: https://commons.wikimedia.org/wiki/File:Total current in phasor domain.svg, CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), https://commons.wikimedia.org/wiki/File:Loss tangent definition.svg, Fig. 3.5: CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/).

47

Chapter 4

Current Flow in Imperfect Conductors 4.1 AC Current Flow in a Good Conductor [m0069]

magnitude of E decreases in proportion to e−αd where α is the attenuation constant and d is distance traversed by the wave. The attenuation constant increases with increasing σ, so the rate of decrease of the magnitude of E increases with increasing σ.

To establish context, consider the simple DC circuit shown in Figure 4.1. In this circuit, the current source provides a steady current which flows through a cylinder-shaped wire. As long as the conductivity σ (SI base units of S/m) of the wire is uniform throughout the wire, the current density J (SI base units of A/m2 ) is uniform throughout the wire.

In the limiting case of a perfect conductor, α → ∞ and so E → 0 everywhere inside the material. Any current within the wire must be the result of either an impressed source or it must be a response to E. Without either of these, we conclude that in the AC case, J → 0 everywhere inside a perfect conductor.1 But if J = 0 in the material, then how does current pass through the wire? We are forced to conclude that the current must exist as a surface current; i.e., entirely outside the wire, yet bound to the surface of the wire. Thus:

Now let us consider the AC case. Whereas the electric field intensity E is constant in the DC case, E exists as a wave in the AC case. In a good conductor, the

In the AC case, the current passed by a perfectlyconducting material lies entirely on the surface of the material.

In this section, we consider the distribution of current in a conductor which is imperfect (i.e., a “good conductor”) and at frequencies greater than DC.

The perfectly-conducting case is unobtainable in practice, but the result gives us a foothold from which we may determine what happens when σ is not infinite. If σ is merely finite, then α is also finite and subsequently wave magnitude may be non-zero over finite distances. Let us now consider the direction in which this putative wave propagates. The two principal directions in the present problem are parallel to the c C. Wang CC BY-SA 4.0

Figure 4.1: Current flow in cylinder at DC.

1 You might be tempted to invoke Ohm’s law (J = σE) to argue against this conclusion. However, Ohm’s law provides no useful information about the current in this case, since σ → ∞ at the same time E → 0. What Ohm’s law is really saying in this case is that E = J/σ → 0 because J must be finite and σ → ∞.

c 2020 S.W. Ellingson CC BY SA 4.0. https://doi.org/10.21061/electromagnetics-vol-2 Electromagnetics Vol. 2.

4.1. AC CURRENT FLOW IN A GOOD CONDUCTOR

49

axis of the wire and perpendicular to the axis of the wire. Waves propagating in any other direction may be expressed as a linear combination of waves traveling in the principal directions, so we need only consider the principal directions to obtain a complete picture. Consider waves propagating in the perpendicular direction first. In this case, we presume a boundary condition in the form of non-zero surface current, which we infer from the perfectly-conducting case considered previously. We also note that E deep inside the wire must be weaker than E closer to the surface, since a wave deep inside the wire must have traversed a larger quantity of material than a wave measured closer to the surface. Applying Ohm’s law (J = σE), the current deep inside the wire must similarly diminish for a good conductor. We conclude that a wave traveling in the perpendicular direction exists and propagates toward the center of the wire, decreasing in magnitude with increasing distance from the surface.

by Biezl (modified) (public domain)

Figure 4.2: Distribution of AC current in a wire of circular cross-section. Shading indicates current density. depends on frequency. In particular, the current will be concentrated close to the surface at high frequencies, uniformly distributed throughout the wire at DC, and in an intermediate state for intermediate frequencies.

We are unable to infer the presence of a wave traveling in the other principal direction – i.e., along Additional Reading: the axis of the wire – since there is no apparent boundary condition to be satisfied on either end of the wire. Furthermore, the presence of such a wave would • “Skin effect” on Wikipedia. mean that different cross-sections of the wire exhibit different radial distributions of current. This is not consistent with physical observations. We conclude that the only relevant wave is one which travels from the surface of the wire inward. Since the current density is proportional to the electric field magnitude, we conclude: In the AC case, the current passed by a wire comprised of a good conductor is distributed with maximum current density on the surface of the wire, and the current density decays exponentially with increasing distance from the surface. This phenomenon is known as the skin effect, referring to the notion of current forming a skin-like layer below the surface of the wire. The effect is illustrated in Figure 4.2. Since α increases with increasing frequency, we see that the specific distribution of current within the wire

CHAPTER 4. CURRENT FLOW IN IMPERFECT CONDUCTORS

50

4.2 Impedance of a Wire [m0159]

The goal of this section is to determine the impedance – the ratio of potential to current – of a wire. The answer to this question is relatively simple in the DC (“steady current”) case: The impedance is found to be equal to the resistance of the wire, which is given by l (DC) (4.1) σA where l is the length of the wire and A is the cross-sectional area of the wire. Also, the impedance of a wire comprised of a perfect conductor at any frequency is simply zero, since there is no mechanism in the wire that can dissipate or store energy in this case. R=

However, all practical wires are comprised of good – not perfect – conductors, and of course many practical signals are time-varying, so the two cases above do not address a broad category of practical interest. The more general case of non-steady currents in imperfect conductors is complicated by the fact that the current in an imperfect conductor is not uniformly distributed in the wire, but rather is concentrated near the surface and decays exponentially with increasing distance from the surface (this is determined in Section 4.1). We are now ready to consider the AC case for a wire comprised of a good but imperfect conductor. What is the impedance of the wire if the current source is sinusoidally-varying? Equation 4.1 for the DC case was determined by first obtaining expressions for the potential (V ) and net current (I) for a length-l section of wire in terms of the electric field intensity E in the wire. To follow that same approach here, we require an expression for I in terms of E that accounts for the non-uniform distribution of current. This is quite difficult to determine in the case of a cylindrical wire. However, we may develop an approximate solution by considering a surface which is not cylindrical, but rather planar. Here we go: Consider the experiment described in Figure 4.3. Here, a semi-infinite region filled with a homogeneous good conductor meets a semi-infinite region of free space along a planar interface at z = 0, with increasing z corresponding to increasing depth

into the material. A plane wave propagates in the +ˆ z direction, beginning from just inside the structure’s surface. The justification for presuming the existence of this wave was presented in Section 4.1. The electric field intensity is given by e =x ˆ E0 e−αz e−jβz E

(4.2)

e = σE e =x ˆ σE0 e−αz e−jβz J

(4.3)

where E0 is an arbitrary complex-valued constant. The current density is given by Ohm’s law of electromagnetics:2

Recall that α = δs−1 where δs is skin depth (see Section 3.12). Also, for a good conductor, we know that β ≈ α (Section 3.11). Using these relationships, we may rewrite Equation 4.3 as follows: e≈x ˆ σE0 e−z/δs e−jz/δs J ˆ σE0 e−(1+j)z/δs =x

(4.4)

e over The net current Ie is obtained by integrating J any cross-section S through which all the current flows; i.e., Z e · ds Ie = J (4.5) S

Here, the simplest solution is obtained by choosing S to be a rectangular surface that is perpendicular to the e at x = 0. This is shown in Figure 4.4. direction of J

c C. Wang CC BY-SA 4.0

Figure 4.3: Experiment used to determine current density and resistance in the AC case.

4.2. IMPEDANCE OF A WIRE

51 where we have determined that x = 0 corresponds to the “+” terminal and x = l corresponds to the “−” terminal.3 The path of integration can be any path that begins and ends at the proscribed terminals. The simplest path to use is one along the surface, parallel to the x axis. Along this path, z = 0 and thus e =x ˆ E0 . For this path: E Ve = −

Z

0

x=l

(ˆ xE0 ) · (ˆ xdx) = E0 l

(4.10)

The impedance Z measured across terminals at x = 0 and x = l is now determined to be: Z, c Y. Zhao CC BY-SA 4.0

Figure 4.4: Choice of S for calculating net current I. The dimensions of S are width W in the y dimension and extending to infinity in the z direction. Then we have Z W Z ∞  ˆ σE0 e−(1+j)z/δs · (ˆ Ie ≈ x x dy dz) y=0

z=0 Z ∞

Ve 1+j l ≈ · e σδs W I

(4.11)

The resistance is simply the real part, so we obtain R≈

l σ(δs W )

(AC case)

(4.12)

The quantity R in this case is referred to specifically as the ohmic resistance, since it is due entirely to the limited conductivity of the material as quantified by Ohm’s law.4

(4.6) Note the resemblance to Equation 4.1 (the solution for the DC case): In the AC case, the product δs W , z=0 having units of area, plays the role of the physical For convenience, let us define the constant cross-section S. Thus, we see an interesting new K , (1 + j)/δs . Since K is constant with respect to interpretation of the skin depth δs : It is the depth to z, the remaining integral is straightforward to which a uniform (DC) current would need to flow in evaluate: order to produce a resistance equal to the observed ∞ Z ∞ (AC) resistance. 1 e−Kz dz = − e−Kz K 0 0 Equations 4.11 and 4.12 were obtained for a good 1 (4.7) conductor filling an infinite half-space, having a flat =+ K surface. How well do these results describe a = σE0 W

e−(1+j)z/δs dz

cylindrical wire? The answer depends on the radius of the wire, a. For δs ≪ a, Equations 4.11 and 4.12 are δs Ie ≈ σE0 W (4.8) excellent approximations, since δs ≪ a implies that 1+j most of the current lies in a thin shell close to the surface of the wire. In this case, the model used to develop the equations is a good approximation for any We calculate Ve for a length l of the wire as follows: given radial slice through the wire, and we are Z 0 3 If this is not clear, recall that the electric field vector must point e · dl Ve = − E (4.9) Incorporating this result into Equation 4.6, we obtain:

x=l

2 To be clear, this is the “point form” of Ohm’s law, as opposed to the circuit theory form (V = IR).

away from positive charge (thus, the + terminal). 4 This is in contrast to other ways that voltage and current can be related; for example, the non-linear V -I characteristic of a diode, which is not governed by Ohm’s law.

CHAPTER 4. CURRENT FLOW IN IMPERFECT CONDUCTORS

52

justified in replacing W with the circumference 2πa. Thus, we obtain the following expressions: Z≈

1+j l · (δs ≪ a) σδs 2πa

The resistance of a wire comprised of a good but imperfect conductor is proportional to the square root of frequency.

(4.13)

At this point, we have determined that resistance is given approximately by Equation 4.17 for δs ≪ a, and so l corresponding to frequencies in the MHz regime and (δs ≪ a) R≈ (4.14) above, and by Equation 4.1 for δs ≫ a, typically σ(δs 2πa) corresponding to frequencies in the kHz regime and below. We have also found that resistance changes √ The impedance of a wire of length l and radius slowly with frequency; i.e., in proportion to f . a ≫ δs is given by Equation 4.13. The resistance Thus, it is often possible to roughly estimate of such a wire is given by Equation 4.14. resistance at frequencies between these two frequency regimes by comparing the DC resistance from Equation 4.1 to the AC resistance from Equation 4.17. If, on the other hand, a < δs or merely ∼ δs , then An example follows. current density is significant throughout the wire, including along the axis of the wire. In this case, we cannot assume that the current density decays Example 4.1. Resistance of the inner conductor smoothly to zero with increasing distance from the of RG-59. surface, and so the model leading to Equation 4.13 is a poor approximation. The frequency required for Elsewhere we have considered RG-59 coaxial validity of Equation 4.13 can be determined by noting cable (see the section “Coaxial Line,” which √ that δs ≈ 1/ πf µσ for a good conductor; therefore, may appear in another volume depending on the we require version of this book). We noted that it was not 1 possible to determine the AC resistance per unit √ ≪a (4.15) πf µσ length R′ for RG-59 from purely electrostatic and magnetostatic considerations. We are now for the derived expressions to be valid. Solving for f , able to consider the resistance per unit length of we find: 1 the inner conductor, which is a solid wire of the f≫ (4.16) πµσa2 type considered in this section. Let us refer to ′ this quantity as Ric . Note that For commonly-encountered wires comprised of typical good conductors, this condition applies at frequencies in the MHz regime and above. These results lead us to one additional interesting finding about the AC resistance of wires. Since √ δs ≈ 1/ πf µσ for a good conductor, Equation 4.14 may be rewritten in the following form: 1 R≈ 2

r

µf l · πσ a

(4.17)

We√have found that R is approximately proportional to f . For example, increasing frequency by a factor of 4 increases resistance by a factor of 2. This frequency dependence is evident in all kinds of practical wires and transmission lines. Summarizing:

′ ′ R′ = Ric + Roc

(4.18)

′ where Roc is the resistance per unit length of the ′ outer conductor. Roc remains a bit too ′ complicated to address here. However, Ric is ′ ′ ′ typically much greater than Roc , so R ∼ Ric . ′ That is, we get a pretty good idea of R for RG-59 by considering the inner conductor alone.

The relevant parameters of the inner conductor are µ ≈ µ0 , σ ∼ = 2.28 × 107 S/m, and

4.2. IMPEDANCE OF A WIRE

53

a∼ = 0.292 mm. Using Equation 4.17, we find: r Ric 1 µf 1 ′ Ric , = · l 2 πσ a  p ∼ f (4.19) = 227 µΩ · m−1 · Hz−1/2

Using Expression 4.16, we find this is valid only for f ≫ 130 kHz. So, for example, we may be ′ confident that Ric ≈ 0.82 Ω/m at 13 MHz.

At the other extreme (f ≪ 130 kHz), Equation 4.1 (the DC resistance) is a better estimate. In this low frequency case, we estimate ′ that Ric ≈ 0.16 Ω/m and is approximately constant with frequency. We now have a complete picture: As frequency is increased from DC to 13 MHz, we expect that ′ will increase monotonically from Ric ≈ 0.16 Ω/m to ≈ 0.82 Ω/m, √ and will continue to increase in proportion to f from that value. Returning to Equation 4.11, we see that resistance is not the whole story here. The impedance Z = R + jX also has a reactive component X equal to the resistance R; i.e., l X≈R≈ σ(δs 2πa)

(4.20)

This is unique to good conductors at AC; that is, we see no such reactance at DC. Because this reactance is positive, it is often referred to as an inductance. However, this is misleading since inductance refers to the ability of a structure to store energy in a magnetic field, and energy storage is decidedly not what is happening here. The similarity to inductance is simply that this reactance is positive, as is the reactance associated with inductance. As long as we keep this in mind, it is reasonable to model the reactance of the wire as an equivalent inductance: Leq ≈

l 1 · 2πf σ(δs 2πa)

(4.21)

Now substituting an expression for skin depth: r 1 πf µ l Leq ≈ · · 2πf σ 2πa r 1 µ l · (4.22) = 4π 3/2 σf a

for a wire having a circular cross-section with δs ≪ a. The utility of this description is that it facilitates the modeling of wire reactance as an inductance in an equivalent circuit. Summarizing: A practical wire may be modeled using an equivalent circuit consisting of an ideal resistor (Equation 4.17) in series with an ideal inductor (Equation 4.22). Whereas resistance increases with the square root of frequency, inductance decreases with the square root of frequency. If the positive reactance of a wire is not due to physical inductance, then to what physical mechanism shall we attribute this effect? A wire has reactance because there is a phase shift between potential and current. This is apparent by comparing Equation 4.6 to Equation 4.10. This is the same phase shift that was found to exist between the electric and magnetic fields propagating in a good conductor, as explained in Section 3.11. Example 4.2. Equivalent inductance of the inner conductor of RG-59. Elsewhere in the book we worked out that the inductance per unit length L′ of RG-59 coaxial cable was about 370 nH/m. We calculated this from magnetostatic considerations, so the reactance associated with skin effect is not included in this estimate. Let’s see how L′ is affected by skin effect for the inner conductor. Using Equation 4.22 with µ = µ0 , σ∼ = 2.28 × 107 S/m, and a ∼ = 0.292 mm, we find  l Leq ≈ 3.61 × 10−5 H·m−1 ·Hz1/2 √ f (4.23) Per unit length: L′eq ,

Leq 3.61 × 10−5 H·Hz1/2 √ ≈ l f

(4.24)

This equals the magnetostatic inductance per unit length (≈ 370 nH/m) at f ≈ 9.52 kHz, and decreases with increasing frequency. Summarizing: The equivalent inductance

CHAPTER 4. CURRENT FLOW IN IMPERFECT CONDUCTORS

54

associated with skin effect is as important as the magnetostatic inductance in the kHz regime, and becomes gradually less important with increasing frequency. Recall that the phase velocity in a low-loss √ transmission line is approximately 1/ L′ C ′ . This means that skin effect causes the phase velocity in such lines to decrease with decreasing frequency. In other words: Skin effect in the conductors comprising common transmission lines leads to a form of dispersion in which higher frequencies travel faster than lower frequencies. This phenomenon is known as chromatic dispersion, or simply “dispersion,” and leads to significant distortion for signals having large bandwidths.

4.3 Surface Impedance [m0160]

In Section 4.2, we derived the following expression for the impedance Z of a good conductor having width W , length l, and which is infinitely deep: Z≈

(AC case)

(4.25)

where σ is conductivity (SI base units of S/m) and δs is skin depth. Note that δs and σ are constitutive parameters of material, and do not depend on geometry; whereas l and W describe geometry. With this in mind, we define the surface impedance ZS as follows: 1+j ZS , (4.26) σδs so that Z ≈ ZS

Additional Reading: • “Skin effect” on Wikipedia.

1+j l · σδs W

l W

(4.27)

Unlike the terminal impedance Z, ZS is strictly a materials property. In this way, it is like the intrinsic or “wave” impedance η, which is also a materials property. Although the units of ZS are those of impedance (i.e., ohms), surface impedance is usually indicated as having units of “Ω/” (“ohms per square”) to prevent confusion with the terminal impedance. Summarizing: Surface impedance ZS (Equation 4.26) is a materials property having units of Ω/, and which characterizes the AC impedance of a material independently of the length and width of the material. Surface impedance is often used to specify sheet materials used in the manufacture of electronic and semiconductor devices, where the real part of the surface impedance is more commonly known as the surface resistance or sheet resistance. Additional Reading: • “Sheet resistance” on Wikipedia. [m0184]

4.3. SURFACE IMPEDANCE

Image Credits c Sevenchw (C. Wang), Fig. 4.1: https://commons.wikimedia.org/wiki/File:Current flow in cylinder new.svg, CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). Fig. 4.2: Biezl, https://commons.wikimedia.org/wiki/File:Skin depth.svg, public domain. Modified from original. c Sevenchw (C. Wang), Fig. 4.3: https://commons.wikimedia.org/wiki/File:Experiment for current density and resistance.svg, CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c YahuiZ (Y. Zhao), https://commons.wikimedia.org/wiki/File:Figure4.4.svg, Fig. 4.4: CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). Modified by author.

55

Chapter 5

Wave Reflection and Transmission 5.1 Plane Waves at Normal Incidence on a Planar Boundary

the associated magnetic field intensity is i

E e i (z) = y ˆ 0 e−jβ1 z , z ≤ 0 H η1 p where η1 = µ1 /ǫ1 is the wave impedance in Region 1.

[m0161]

When a plane wave encounters a discontinuity in media, reflection from the discontinuity and transmission into the second medium is possible. In this section, we consider the scenario illustrated in Figure 5.1: a uniform plane wave which is normally incident on the planar boundary between two semi-infinite material regions. By “normally-incident” we mean the direction of ˆ is perpendicular to the boundary. We propagation k shall assume that the media are “simple” and lossless (i.e., the imaginary component of permittivity ǫ′′ is equal to zero) and therefore the media are completely defined by a real-valued permittivity and a real-valued permeability.

(5.2)

The possibility of a reflected plane wave propagating in exactly the opposite direction is inferred from two pieces of evidence: The general solution to the wave equation, which includes terms corresponding to waves traveling in both +ˆ z or −ˆ z; and the geometrical symmetry of the problem, which precludes waves traveling in any other directions. The symmetry of the problem also precludes a change of ˆ polarization, so the reflected wave should have no y component. Therefore, we may be confident that the reflected electric field has the form e r (z) = x ˆ Be+jβ1 z , z ≤ 0 E

(5.3)

Figure 5.1 shows the wave incident on the boundary, which is located at the z = 0 plane. The electric field e i of this wave is given by intensity E e i (z) = x ˆ E0i e−jβ1 z , z ≤ 0 E

(5.1)

√ where β1 = ω µ1 ǫ1 is the phase propagation constant in Region 1 and E0i is a complex-valued e i serves as the “stimulus” in this problem. constant. E That is, all other contributions to the total field may e i . In fact, all other be expressed in terms of E contributions to the total field may be expressed in terms of E0i . From the plane wave relationships, we determine that

c C. Wang CC BY-SA 4.0

Figure 5.1: A uniform plane wave normally incident on the planar boundary between two semi-infinite material regions.

c 2020 S.W. Ellingson CC BY SA 4.0. https://doi.org/10.21061/electromagnetics-vol-2 Electromagnetics Vol. 2.

5.1. PLANE WAVES AT NORMAL INCIDENCE ON A PLANAR BOUNDARY where B is a complex-valued constant that remains to be determined. Since the direction of propagation for the reflected wave is −ˆ z, we have from the plane wave relationships that

57

Now employing Equations 5.1, 5.3, and 5.5, we obtain: E0i + B = C (5.10)

Clearly a second equation is required to determine (5.4) both B and C. This equation can be obtained by enforcing the boundary condition on the magnetic field. Recall that any discontinuity in the tangential Similarly, we infer the existence of a “transmitted” component of the total magnetic field intensity must plane wave propagating on the z > 0 side of the be supported by a current flowing on the surface. boundary. The symmetry of the problem precludes There is no impressed current in this problem, and any direction of propagation other than +z, and with there is no reason to suspect a current will arise in no possibility of a wave traveling in the −ˆ z direction response to the fields present in the problem. for z > 0. Therefore, we may be confident that the Therefore, the tangential components of the magnetic transmitted electric field has the form: field must be continuous across the boundary. This e t (z) = x ˆ Ce−jβ2 z , z ≥ 0 E (5.5) becomes the same boundary condition that we applied to the total electric field intensity, so the remaining e 1 and H e 2 to be the and an associated magnetic field having the form steps are the same. We define H total magnetic field intensities in Regions 1 and 2, C e t (z) = y ˆ e−jβ2 z , z ≥ 0 H (5.6) respectively. The total field in Region 1 is η2 p e 1 (z) = H e i (z) + H e r (z) √ H (5.11) where β2 = ω µ2 ǫ2 and η2 = µ2 /ǫ2 are the phase propagation constant and wave impedance, The field in Region 2 is simply respectively, in Region 2. The constant C, like B, is a complex-valued constant that remains to be e 2 (z) = H e t (z) H (5.12) determined. e 1 (0) = H e 2 (0), The boundary condition requires H At this point, the only unknowns in this problem are and therefore the complex-valued constants B and C. Once these values are known, the problem is completely solved. e i (0) + H e r (0) = H e t (0) H (5.13) These values can be determined by the application of boundary conditions at z = 0. First, recall that the Now employing Equations 5.2, 5.4, and 5.6, we tangential component of the total electric field obtain: B C E0i intensity must be continuous across a material − = (5.14) boundary. To apply this boundary condition, let us η1 η1 η2 e 1 and E e 2 to be the total electric field define E intensities in Regions 1 and 2, respectively. The total Equations 5.10 and 5.14 constitute a linear system of field in Region 1 is the sum of incident and reflected two simultaneous equations with two unknowns. A fields, so straightforward method of solution is to first eliminate e 1 (z) = E e i (z) + E e r (z) E (5.7) C by substituting the left side of Equation 5.10 into Equation 5.14, and then to solve for B. One obtains: The field in Region 2 is simply B e r (z) = −ˆ H y e+jβ1 z , z ≤ 0 η1

e 2 (z) = E e t (z) E

Also, we note that all field components are already tangent to the boundary. Thus, continuity of the tangential component of the electric field across the e 1 (0) = E e 2 (0), and therefore boundary requires E e i (0) + E e r (0) = E e t (0) E

B = Γ12 E0i

(5.8)

(5.15)

where Γ12 ,

η2 − η1 η2 + η1

(5.16)

Γ12 is known as a reflection coefficient. The subscript (5.9) “12” indicates that this coefficient applies for

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

58

incidence from Region 1 toward Region 2. We may now solve for C by substituting Equation 5.15 into Equation 5.10. We find: C = (1 + Γ12 ) E0i Now summarizing the solution: e r (z) = x ˆ Γ12 E0i e+jβ1 z , z ≤ 0 E

e t (z) = x ˆ (1 + Γ12 ) E0i e−jβ2 z , z ≥ 0 E

(5.17) It may be helpful to note the very strong analogy between reflection of the electric field component of a plane wave from a planar boundary and the reflection (5.18) of a voltage wave in a transmission line from a terminating impedance. In a transmission line, the voltage reflection coefficient Γ is given by (5.19)

Equations 5.18 and 5.19 are the reflected and transmitted fields, respectively, in response to the incident field given in Equation 5.1 in the normal incidence scenario shown in Figure 5.1.

e r and H e t may be obtained by Expressions for H applying the plane wave relationships to the preceding expressions.

When Region 2 is a perfect conductor, the reflection coefficient Γ12 = −1 and the solution described in Equations 5.18 and 5.19 applies.

Γ=

ZL − Z0 ZL + Z0

(5.20)

where ZL is the load impedance and Z0 is the characteristic impedance of the transmission line. Comparing this to Equation 5.16, we see η1 is analogous to Z0 and η2 is analogous to ZL . Furthermore, we see the special case η2 = η1 , considered previously, is analogous to a matched load, and the special case η2 = 0, also considered previously, is analogous to a short-circuit load.

In the case of transmission lines, we were concerned about what fraction of the power was delivered into a It is useful to check this solution by examining some load and what fraction of the power was reflected special cases. First: If the material in Region 2 is from a load. A similar question applies in the case of identical to the material in Region 1, then there should plane wave reflection. In this case, we are concerned et = E e i . In this case, η2 = η1 , be no reflection and E about what fraction of the power density is so Γ12 = 0, and we obtain the expected result. transmitted into Region 2 and what fraction of the power density is reflected from the boundary. The A second case of practical interest is when Region 2 i associated with the time-average power density Save is a perfect conductor. First, note that this may seem incident wave is: at first glance to be a violation of the “lossless” assumption made at the beginning of this section. |E i |2 i Save = 0 (5.21) While it is true that we did not explicitly account for 2η1 the possibility of a perfect conductor in Region 2, let’s see what the present analysis has to say about presuming that E0i is expressed in peak (as opposed to this case. If the material in Region 2 is a perfect rms) units. Similarly, the time-average power density conductor, then there should be no transmission since S r associated with the reflected wave is: ave the electric field is zero in a perfect conductor. In this case, η2 = 0 since the ratio of electric field intensity Γ12 E0i 2 r Save = to magnetic field intensity is zero in Region 2, and 2η1 subsequently Γ12 = −1 and 1 + Γ12 = 0. As i 2 t e is found to be zero, and the reflected expected, E 2 E0 = |Γ12 | electric field experiences a sign change as required to 2η1 e i (0) + E e r (0) = 0. 2 i enforce the boundary condition E = |Γ12 | Save (5.22) Thus, we obtained the correct answer because we were able to independently determine that η2 = 0 in a From the principle of conservation of power, the t perfect conductor. power density Save transmitted into Region 2 must be

5.1. PLANE WAVES AT NORMAL INCIDENCE ON A PLANAR BOUNDARY equal to the incident power density minus the reflected power density. Thus: t i r Save = Save − Save  i = 1 − |Γ12 |2 Save

59

as a low-loss dielectric with relative permittivity of about 3. Characterize the efficiency of reflection of a radio wave from the Moon. (5.23)

In other words, the ratio of power density transmitted into Region 2 to power density incident from Region 1 is t Save (5.24) = 1 − |Γ12 |2 i Save Again, the analogy to transmission line theory is evident. The fractions of power density reflected and transmitted relative to power density incident are given in terms of the reflection coefficient Γ12 by Equations 5.22 and 5.24, respectively. Finally, note that a variety of alternative expressions exist for the reflection coefficient Γ12 . Since the wave p impedance η = µ/ǫ in lossless media, and since most lossless materials are non-magnetic (i.e., exhibit µ ≈ µ0 ), it is possible to express Γ12 purely in terms of permittivity for these media. Assuming µ = µ0 , we find: η2 − η1 Γ12 = η2 + η1 p p µ0 /ǫ2 − µ0 /ǫ1 p =p µ0 /ǫ2 + µ0 /ǫ1 √ √ ǫ1 − ǫ2 (5.25) =√ √ ǫ1 + ǫ2

Solution. At radio frequencies, a wavelength is many orders of magnitude less than the diameter of the Moon, so we are justified in treating the Moon’s surface as a planar boundary between free space and a semi-infinite region of lossy dielectric material. The reflection coefficient is given approximately by Equation 5.26 with ǫr1 ≈ 1 and ǫr2 ∼ 3. Thus, Γ12 ∼ −0.27. Subsequently, the fraction of power reflected from the Moon relative to power incident is 2 |Γ12 | ∼ 0.07; i.e., about 7%. In optics, it is common to make the definition √ n , ǫr where n is known as the index of refraction or refractive index.1 In terms of the indices of refraction n1 and n2 for Regions 1 and 2, respectively: n1 − n2 Γ12 = (5.27) n1 + n2 Additional Reading: • “Refractive index” on Wikipedia.

Moreover, recall that permittivity can be expressed in terms of relative permittivity; that is, ǫ = ǫr ǫ0 . Making the substitution above and eliminating all extraneous factors of ǫ0 , we find: √ √ ǫr1 − ǫr2 Γ12 = √ (5.26) √ ǫr1 + ǫr2 where ǫr1 and ǫr2 are the relative permittivities in Regions 1 and 2, respectively. Example 5.1. Radio reflection from the surface of the Moon. At radio frequencies, the Moon can be modeled

1 A bit of a misnomer, since the definition applies even when there is no refraction, as in the scenario considered in this section.

60

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

5.2 Plane Waves at Normal Incidence on a Material Slab [m0162]

In Section 5.1, we considered what happens when a uniform plane wave is normally incident on the planar boundary between two semi-infinite media. In this section, we consider the problem shown in Figure 5.2: a uniform plane wave normally incident on a “slab” sandwiched between two semi-infinite media. This scenario arises in many practical engineering problems, including the design and analysis of filters and impedance-matching devices at RF and optical frequencies, the analysis of RF propagation through walls, and the design and analysis of radomes. The findings of Section 5.1 are an important stepping stone to the solution of this problem, so a review of Section 5.1 is recommended before reading this section. For consistency of terminology, let us refer to the problem considered in Section 5.1 as the “single-boundary” problem and the present (slab) problem as the “double-boundary” problem. Whereas there are only two regions (“Region 1” and “Region 2”) in the single-boundary problem, in the double-boundary problem there is a third region that we shall refer to as “Region 3.” We assume that the media comprising each slab are “simple” and lossless (i.e., the imaginary component of permittivity ǫ′′ is equal to zero) and therefore the media are completely defined by a real-valued permittivity and a real-valued

permeability. The boundary between Regions 1 and 2 is at z = −d, and the boundary between Regions 2 and 3 is at z = 0. Thus, the thickness of the slab is d. In both problems, we presume an incident wave in Region 1 incident on the boundary with Region 2 having electric field intensity e i (z) = x ˆ E0i e−jβ1 z (Region 1) E

(5.28)

√ where β1 = ω µ1 ǫ1 is the phase propagation e i serves as the “stimulus” in constant in Region 1. E this problem. That is, all other contributions to the e i. total field may be expressed in terms of E

From the plane wave relationships, we determine that the associated magnetic field intensity is i

E e i (z) = y ˆ 0 e−jβ1 z (Region 1) (5.29) H η1 p where η1 = µ1 /ǫ1 is the wave impedance in Region 1. The symmetry arguments of the single-boundary problem apply in precisely the same way to the double-boundary problem. Therefore, we presume that the reflected electric field intensity is: e r (z) = x ˆ Be+jβ1 z (Region 1) E

(5.30)

where B is a complex-valued constant that remains to be determined; and subsequently the reflected magnetic field intensity is: B e r (z) = −ˆ H y e+jβ1 z (Region 1) η1

(5.31)

Similarly, we infer the existence of a transmitted plane wave propagating in the +ˆ z direction in Region 2. The electric and magnetic field intensities of this wave are given by: e t2 (z) = x ˆ Ce−jβ2 z (Region 2) E

(5.32)

and an associated magnetic field having the form: C e t2 (z) = y ˆ e−jβ2 z (Region 2) H (5.33) η2 c C. Wang CC BY-SA 4.0

p √ where β2 = ω µ2 ǫ2 and η2 = µ2 /ǫ2 are the phase Figure 5.2: A uniform plane wave normally incident propagation constant and wave impedance, on a slab. respectively, in Region 2. The constant C, like B, is a

5.2. PLANE WAVES AT NORMAL INCIDENCE ON A MATERIAL SLAB complex-valued constant that remains to be determined.

condition is that the total field should be continuous across each boundary. Applying this condition to the boundary at z = 0, we obtain:

Now let us consider the boundary between Regions 2 e t2 is incident on this boundary in and 3. Note that E e i is incident on the precisely the same manner as E boundary between Regions 1 and 2. Therefore, we infer a reflected electric field intensity in Region 2 as follows: e r2 (z) = x ˆ De+jβ2 z (Region 2) E

(5.34)

where D is a complex-valued constant that remains to be determined; and an associated magnetic field D e r2 (z) = −ˆ H y e+jβ2 z (Region 2) η2

e t2 (0) + E e r2 (0) = E e t (0) E

e t2 (0) + H e r2 (0) = H e t (0) H

(5.38) (5.39)

Applying this condition to the boundary at z = −d, we obtain: e i (−d) + E e r (−d) = E e t2 (−d) + E e r2 (−d) (5.40) E

e i (−d) + H e r (−d) = E e t2 (−d) + E e r2 (−d) (5.41) H

Making substitutions from Table 5.1 and dividing out (5.35) common factors, Equation 5.38 becomes:

Subsequently, we infer a transmitted electric field intensity in Region 3 as follows: e t (z) = x ˆ F e−jβ3 z (Region 3) E

61

(5.36)

and an associated magnetic field having the form

C +D =F

(5.42)

Equation 5.39 becomes: D F C − = η2 η2 η3

(5.43)

Equation 5.40 becomes: F e t (z) = y ˆ e−jβ3 z (Region 3) H (5.37) η3 E0i e+jβ1 d + Be−jβ1 d = Ce+jβ2 d + De−jβ2 d (5.44) p √ where β3 = ω µ3 ǫ3 and η3 = µ3 /ǫ3 are the phase Equation 5.41 becomes: propagation constant and wave impedance, respectively, in Region 3. The constant F , like D, is a E0i +jβ1 d B −jβ1 d C D complex-valued constant that remains to be e − e = e+jβ2 d − e−jβ2 d η1 η1 η2 η2 determined. We infer no wave traveling in the (−ˆ z) in (5.45) Region 3, just as we inferred no such wave in Region 2 of the single-boundary problem. Equations 5.42–5.45 are recognizable as a system of 4 simultaneous linear equations with the number of For convenience, Table 5.1 shows a complete unknowns equal to the number of equations. We summary of the field components we have just could simply leave it at that, however, some very identified. Note that the total field intensity in each useful insights are gained by solving this system of region is the sum of the field components in that equations in a particular manner. First, note the region. For example, the total electric field intensity resemblance between the situation at the z = 0 t2 r2 e +E e . in Region 2 is E boundary in this problem and the situation at z = 0 boundary in the single-boundary problem. In fact, the Now observe that there are four unknown constants two are the same problem, with the following remaining; namely B, C, D, and F . The transformation of variables (single-boundary → double-boundary problem is completely solved once double-boundary): we have expressions for these constants in terms of the “given” quantities in the problem statement. Solutions for the unknown constants can be obtained by enforcing boundary conditions on the electric and magnetic fields at each of the two boundaries. As in the single-boundary case, the relevant boundary

E0i → C i.e., the independent variable

(5.46)

B → D i.e., reflection

(5.47)

C → F i.e., transmission

(5.48)

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

62

Region 1 Region 2 Region 3

Electric Field Intensity e i (z) = x ˆ E0i e−jβ1 z E e r (z) = x ˆ Be+jβ1 z E t2 e ˆ Ce−jβ2 z E (z) = x r2 e (z) = x ˆ De+jβ2 z E e t (z) = x ˆ F e−jβ3 z E

Magnetic Field Intensity  e i (z) = +ˆ H y E0i /η1 e−jβ1 z e r (z) = −ˆ H y (B/η1 ) e+jβ1 z t2 e H (z) = +ˆ y (C/η2 ) e−jβ2 z r2 e (z) = −ˆ H y (D/η2 ) e+jβ2 z e t (z) = +ˆ H y (F/η3 ) e−jβ3 z

Region of Validity z ≤ −d −d ≤ z ≤ 0 z≥0

Table 5.1: Field components in the double-boundary (slab) problem of Figure 5.2. It immediately follows that D = Γ23 C

(5.49)

F = (1 + Γ23 ) C

(5.50)

and

where Γ23 ,

η3 − η2 η3 + η2

(5.51)

At this point, we could use the expressions we have just derived to eliminate D and F in the system of four simultaneous equations identified earlier. This would reduce the problem to that of two equations in two unknowns – a dramatic simplification! However, even this is more work than is necessary, and a little cleverness at this point pays big dividends later. The key idea is that we usually have no interest in the fields internal to the slab; in most problems, we are interested merely in reflection into Region 1 from the z = −d boundary and transmission into Region 3 through the z = 0 interface. With this in mind, let us simply replace the two-interface problem in Figure 5.2 with an “equivalent” single-boundary problem shown in Figure 5.3. This problem is “equivalent” in the following sense only: Fields in Region 1 are identical to those in Region 1 of the original problem. The material properties in the region to the right of the z = −d boundary in the equivalent problem seem unlikely to be equal to those in Regions 2 or 3 of the original problem, so we define a new wave impedance ηeq to represent this new condition. If we can find an expression for ηeq , then we can develop a solution to the original (two-boundary) problem that looks like a solution to the equivalent (simpler, single-boundary) problem.

c C. Wang CC BY-SA 4.0

Figure 5.3: Representing the double-boundary problem as an equivalent single-boundary problem. To obtain an expression for ηeq , we invoke the definition of wave impedance: It is simply the ratio of electric field intensity to magnetic field intensity in the medium. Thus: h i e t2 (z2 ) + E e r2 (z2 ) ˆ· E x e 2 (z2 ) ˆ·E x h i ηeq , = e 2 (z2 ) e t2 (z2 ) + H e r2 (z2 ) ˆ·H y ˆ· H y (5.52) where z2 is any point in Region 2. For simplicity, let us choose z2 = −d. Making substitutions, we obtain: ηeq =

Ce+jβ2 d + De−jβ2 d (C/η2 ) e+jβ2 d − (D/η2 ) e−jβ2 d

(5.53)

Bringing the factor of η2 to the front and substituting D = Γ23 C: ηeq = η2

Ce+jβ2 d + Γ23 Ce−jβ2 d Ce+jβ2 d − Γ23 Ce−jβ2 d

(5.54)

Finally, we divide out the common factor of C and multiply numerator and denominator by e−jβ2 d ,

5.2. PLANE WAVES AT NORMAL INCIDENCE ON A MATERIAL SLAB

63

yielding: ηeq

1 + Γ23 e−j2β2 d = η2 1 − Γ23 e−j2β2 d

(5.55)

Equation 5.55 is the wave impedance in the region to the right of the boundary in the equivalent scenario shown in Figure 5.3. “Equivalent” in this case means that the incident and reflected fields in Region 1 are identical to those in the original (slab) problem. Two comments on this expression before proceeding. First: Note that if the materials in Regions 2 and 3 are identical, then η2 = η3 , so Γ23 = 0, and thus ηeq = η2 , as expected. Second: Note that ηeq is, in general, complex-valued. This may initially seem troubling, since the imaginary component of the wave impedance is normally associated with general (e.g., possibly lossy) material. We specifically precluded this possibility in the problem statement, so clearly the complex value of the wave impedance is not indicating loss. Instead, the non-zero phase of ηeq represents the ability of the standing wave inside the slab to impart a phase shift between the electric and magnetic fields. This is precisely the same effect that one observes at the input of a transmission line: The input impedance Zin is, in general, complex-valued even if the line is lossless and the characteristic impedance and load impedance are real-valued.2 In fact, the impedance looking into a transmission line is given by an expression of precisely the same form as Equation 5.55. This striking analogy between plane waves at planar boundaries and voltage and current waves in transmission lines applies broadly. We can now identify an “equivalent reflection coefficient” Γ1,eq for the scenario shown in Figure 5.3: ηeq − η1 (5.56) Γ1,eq , ηeq + η1 The quantity Γ1,eq may now be used precisely in the same way as Γ12 was used in the single-boundary problem to find the reflected fields and reflected power density in Region 1. 2 See the section “Input Impedance of a Terminated Lossless Transmission Line” for a reminder. This section may appear in a different volume depending on the version of this book.

Example 5.2. Reflection of WiFi from a glass pane. A WiFi (wireless LAN) signal at a center frequency of 2.45 GHz is normally incident on a glass pane which is 1 cm thick and is well-characterized in this application as a lossless dielectric with ǫr = 4. The source of the signal is sufficiently distant from the window that the incident signal is well-approximated as a plane wave. Determine the fraction of power reflected from the pane. Solution. In this case, we identify Regions 1 and 3 as approximately free space, and Region 2 as the pane. Thus η1 = η3 = η0 ∼ = 376.7 Ω and

η0 η2 = √ ∼ = 188.4 Ω ǫr

(5.57) (5.58)

The reflection coefficient from Region 2 to Region 3 is η3 − η2 η3 + η2 η0 − η2 = η0 + η2 ∼ = 0.3333

Γ23 =

(5.59)

Given f = 2.45 GHz, the phase propagation constant in the glass is √ 2πf ǫr ∼ 2π = β2 = = 102.6 rad/m (5.60) λ c Given d = 1 cm, the equivalent wave impedance is 1 + Γ23 e−j2β2 d 1 − Γ23 e−j2β2 d ∼ = 117.9 − j78.4 Ω

ηeq = η2

(5.61)

Next we calculate Γ1,eq =

ηeq − η1 ηeq + η1

∼ = −0.4859 − j0.2354

(5.62)

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

64

The ratio of reflected power density to incident power density is simply the squared magnitude of this reflection coefficient, i.e.: r Save 2 = |Γ1,eq | = 0.292 ∼ = 29.2% i Save

5.3 Total Transmission Through a Slab

(5.63) [m0163]

r i where Save and Save are the reflected and incident power densities, respectively.

Section 5.2 details the solution of the “single-slab” problem. This problem is comprised of three material regions: A semi-infinite Region 1, from which a uniform plane wave is normally incident; Region 2, Since B = Γ1,eq E0i , we have: the slab, defined by parallel planar boundaries e r (z) = x ˆ Γ1,eq E0i e+jβ1 z , z ≤ −d E (5.64) separated by distance d; and a semi-infinite Region 3, through which the plane wave exits. The solution that e r (z) can be obtained from the plane wave and H was developed assumes simple media with negligible relationships. If desired, it is now quite simple to loss, so that the media in each region is characterized obtain solutions for the electric and magnetic fields in entirely by permittivity and permeability. Regions 2 and 3. However, it is the usually the power We now focus on a particular class of applications density transmitted into Region 3 that is of greatest interest. This power density is easily determined from involving this structure. In this class of applications, we seek total transmission through the slab. By “total the principle of conservation of power. If the loss in transmission” we mean 100% of power incident on Region 2 is negligible, then no power can be the slab is transmitted through the slab into Region 3, dissipated there. In this case, all power not reflected and 0% of the power is reflected back into Region 1. from the z = −d interface must be transmitted into There are many applications for such a structure. One Region 3. In other words: application is the radome, a protective covering which t partially or completely surrounds an antenna, but Save = 1 − |Γ1,eq |2 (5.65) nominally does not interfere with waves being i Save received by or transmitted from the antenna. Another application is RF and optical wave filtering; that is, t where Save is the transmitted power density. passing or rejecting waves falling within a narrow range of frequencies. In this section, we will first describe the conditions for total transmission, and Example 5.3. Transmission of WiFi through a then shall provide some examples of these glass pane. applications. Continuing Example 5.2: What fraction of We begin with characterization of the media. incident power passes completely through the Region 1 is characterized by its permittivity ǫ1 and glass pane? permeability µ1 , such p that the wave impedance in µ1 /ǫ1 . Similarly, Region 2 is Region 1 is η = 1 Solution. characterized by its permittivity ǫ2 and permeability t Save µ2 , such p that the wave impedance in Region 2 is = 1 − |Γ1,eq |2 ∼ (5.66) = 70.8% i Save η2 = µ2 /ǫ2 . Region 3 is characterized by its permittivity ǫ3 and permeability µ3 , such p that the wave impedance in Region 3 is η3 = µ3 /ǫ3 . The Additional Reading: analysis in this section also depends on β2 , the phase propagation constant in Region 2, which is given by √ ω µ2 ǫ2 . • “Radome” on Wikipedia. Recall that reflection from the slab is quantified by

5.3. TOTAL TRANSMISSION THROUGH A SLAB the reflection coefficient

known as half-wave matching and quarter-wave matching.

ηeq − η1 ηeq + η1

(5.67)

1 + Γ23 e−j2β2 d 1 − Γ23 e−j2β2 d

(5.68)

Γ1,eq = where ηeq is given by ηeq = η2

and where Γ23 is given by Γ23

η3 − η2 = η3 + η2

(5.69)

Total transmission requires that Γ1,eq = 0. From Equation 5.67 we see that Γ1,eq is zero when η1 = ηeq . Now employing Equation 5.68, we see that total transmission requires: η1 = η2

1 + Γ23 e−j2β2 d 1 − Γ23 e−j2β2 d

(5.70)

For convenience and clarity, let us define the quantity P , e−j2β2 d

(5.71)

Using this definition, Equation 5.70 becomes 1 + Γ23 P η1 = η2 1 − Γ23 P

(5.72)

Also, let us substitute Equation 5.69 for Γ23 , and multiply numerator and denominator by η3 + η2 (the denominator of Equation 5.69). We obtain: η1 = η2

η3 + η2 + (η3 − η2 )P η3 + η2 − (η3 − η2 )P

(5.73)

Rearranging the numerator and denominator, we obtain: η1 = η2

(1 + P )η3 + (1 − P )η2 (1 − P )η3 + (1 + P )η2

65

(5.74)

The parameters η1 , η2 , η3 , β2 , and d defining any single-slab structure that exhibits total transmission must satisfy Equation 5.74. Our challenge now is to identify combinations of parameters that satisfy this condition. There are two general categories of solutions. These categories are

Half-wave matching applies when we have the same material on either side of the slab; i.e., η1 = η3 . Let us refer to this common value of η1 and η3 as ηext . Then the condition for total transmission becomes: (1 + P )ηext + (1 − P )η2 ηext = η2 (5.75) (1 − P )ηext + (1 + P )η2 For the above condition to be satisfied, we need the fraction on the right side of the equation to be equal to ηext /η2 . From Equation 5.71, the magnitude of P is always 1, so the first value of P you might think to try is P = +1. In fact, this value satisfies Equation 5.75. Therefore, e−j2β2 d = +1. This new condition is satisfied when 2β2 d = 2πm, where m = 1, 2, 3, ... (We do not consider m ≤ 0 to be valid solutions since these would represent zero or negative values of d.) Thus, we find d=

πm λ2 m , where m = 1, 2, 3, ... = β2 2

(5.76)

where λ2 = 2π/β2 is the wavelength inside the slab. Summarizing: Total transmission through a slab embedded in regions of material having equal wave impedance (i.e., η1 = η3 ) may be achieved by setting the thickness of the slab equal to an integer number of half-wavelengths at the frequency of interest. This is known as half-wave matching. A remarkable feature of half-wave matching is that there is no restriction on the permittivity or permeability of the slab, and the only constraint on the media in Regions 1 and 3 is that they have equal wave impedance. Example 5.4. Radome design by half-wave matching. The antenna for a 60 GHz radar is to be protected from weather by a radome panel positioned directly in front of the radar. The panel is to be constructed from a low-loss material having µr ≈ 1 and ǫr = 4. To have sufficient mechanical integrity, the panel must be at least 3 mm thick. What thickness should be

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

66 used?

Solution. This is a good application for half-wave matching because the material on either side of the slab is the same (presumably free space) whereas the material used for the slab is unspecified. The phase velocity in the slab is c vp = √ ∼ = 1.5 × 108 m/s ǫr

(5.77)

so the wavelength in the slab is λ2 =

vp ∼ = 2.5 mm f

(5.78)

Thus, the minimum thickness of the slab that satisfies the half-wave matching condition is d = λ2 /2 ∼ = 1.25 mm. However, this does not meet the 3 mm minimum-thickness requirement. Neither does the next available thickness, d = λ2 ∼ = 2.5 mm. The next-thickest option, d = 3λ2 /2 ∼ = 3.75 mm, does meet the requirement. Therefore, we select d ∼ = 3.75 mm. It should be emphasized that designs employing half-wave matching will be narrowband – that is, total only for the design frequency. As frequency increases or decreases from the design frequency, there will be increasing reflection and decreasing transmission. Quarter-wave matching requires that the wave impedances in each region are different and related in a particular way. The quarter-wave solution is obtained by requiring P = −1, so that η1 = η2

η2 (1 + P )η3 + (1 − P )η2 = η2 (1 − P )η3 + (1 + P )η2 η3

Note that P = −1 is obtained when 2β2 d = π + 2πm where m = 0, 1, 2, .... Thus, we find d=

Summarizing:

Example 5.5. Radome design by quarter-wave matching. The antenna for a 60 GHz radar is to be protected from weather by a radome panel positioned directly in front of the radar. In this case, however, the antenna is embedded in a lossless material having µr ≈ 1 and ǫr = 2, and the radome panel is to be placed between this material and the outside, which we presume is free space. The material in which the antenna is embedded, and against which the radome panel is installed, is quite rigid so there is no minimum thickness requirement. However, the radome panel must be made from a material which is lossless and non-magnetic. Design the radome panel. Solution. The radome panel must be comprised of a material having r η0 √ η2 = η1 η3 = √ · η0 ∼ = 317 Ω (5.82) 2 Since the radome panel is required to be non-magnetic, the relative permittivity must be given by

(5.79)

Solving this equation for wave impedance of the slab material, we find √ η2 = η1 η3 (5.80)

π π + m 2β2 β2 λ2 λ2 = + m , where m = 0, 1, 2, ... 4 2

Total transmission is achieved through a slab by √ selecting η2 = η1 η3 and making the slab onequarter wavelength at the frequency of interest, or some integer number of half-wavelengths thicker if needed. This is known as quarter-wave matching.

(5.81)

η0 ⇒ ǫr = η2 = √ ǫr



η0 η2

2

∼ = 1.41

(5.83)

The phase velocity in the slab will be c 3 × 108 m/s ∼ √ vp = √ ∼ = = 2.53 × 108 m/s ǫr 1.41 (5.84) so the wavelength in the slab is λ2 =

vp ∼ 2.53 × 108 m/s ∼ = = 4.20 mm (5.85) f 60 × 109 Hz

5.4. PROPAGATION OF A UNIFORM PLANE WAVE IN AN ARBITRARY DIRECTION Thus, the minimum possible thickness of the radome panel is d = λ2 /4 ∼ = 1.05 mm, and the relative permittivity of the radome panel must be ǫr ∼ = 1.41. Additional Reading: • “Radome” on Wikipedia.

67

5.4 Propagation of a Uniform Plane Wave in an Arbitrary Direction [m0165]

An example of a uniform plane wave propagating in a lossless medium is shown in Figure 5.4. This wave is expressed in the indicated coordinate system as follows: e =x ˆ E0 e−jβz E (5.86) This is a phasor-domain expression for the electric field intensity, so E0 is a complex-valued number representing the magnitude and reference phase of the sinusoidally-varying wave. The term “reference e at the origin of the phase” is defined as the phase of E coordinate system. Since the phase propagation constant β is real and positive, this wave is traveling in the +ˆ z direction through simple lossless media.

Note that electric field intensity vector is linearly ˆ . Depending on polarized in a direction parallel to x the position in space (and, for the physical e points either in the time-domain waveform, time), E +ˆ x direction or the −ˆ x direction. Let us be a bit more e To do specific about the direction of the vector E. this, let us define the reference polarization to be the e direction in which  E points when −jβz e is Re E0 e ≥ 0; i.e., when the phase of E between −π/2 and +π/2 radians. Thus, the reference e in Equation 5.86 is always +ˆ polarization of E x.

Note that Equation 5.86 indicates a specific combination of reference polarization and direction of propagation. However, we may obtain any other combination of reference polarization and direction of

c C. Wang CC BY-SA 4.0

Figure 5.4: tion 5.86.

The plane wave described by Equa-

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

68

c C. Wang CC BY-SA 4.0

c C. Wang CC BY-SA 4.0

Figure 5.7: A plane wave described in ray-fixed coordinates, yielding Equation 5.89.

Figure 5.5: The same plane wave described in a rotated coordinate system, yielding Equation 5.87. different due to our choice for the orientation of the coordinate system in each case. Now consider the same thought experiment for the infinite number of cases in which the wave does not propagate in one of the three basis directions of the Cartesian coordinate system. One situation in which we face this complication is when the wave is obliquely incident on a surface. In this case, it is c

C. Wang CC BY-SA 4.0 impossible to select a single orientation of the Figure 5.6: The same plane wave described in yet an- coordinate system in which the directions of other rotation of the coordinate system, yielding Equa- propagation, reference polarization, and surface tion 5.88. normal can all be described in terms of one basis vector each. To be clear: There is no fundamental limitation imposed by this slipperiness of the propagation by rotation of the Cartesian coordinate coordinate system. However, practical problems such system. For example, if we rotate the +x axis of the as the oblique-incidence scenario described above are coordinate system into the position originally much easier to analyze if we are able to express occupied by the +y axis, then the very same wave is waves in a system of coordinates which always yields expressed as the same expressions. e = −ˆ E yE0 e−jβz (5.87) Fortunately, this is easily accomplished using This is illustrated in Figure 5.5. At first glance, it ray-fixed coordinates. Ray-fixed coordinates are a appears that the reference polarization has changed; unique set of coordinates that are determined from the however, this is due entirely to our choice of characteristics of the wave, as opposed to being coordinate system. That is, the reference polarization determined arbitrarily and separately from the is precisely the same; it is only the coordinate system characteristics of the wave. The ray-fixed used to describe the reference polarization that has representation of a uniform plane wave is: changed. e ˆE0 e−jk·r (5.89) E(r) =e Let us now rotate the +z axis of the coordinate system around the x axis into the position originally occupied by the −z axis. Now the very same wave is This is illustrated in Figure 5.7. In this representation, e is evaluated, e ˆ is the r is the position at which E expressed as reference polarization expressed as a unit vector, and e = +ˆ E yE0 e+jβz (5.88) ˆ in the direction of propagation, k is the unit vector k times β; i.e.: This is illustrated in Figure 5.6. At first glance, it ˆ k , kβ (5.90) appears that the direction of propagation has reversed; but, again, it is only the coordinate system that has Consider Equation 5.86 as an example. In this case, changed, and not the direction of propagation. ˆ=x ˆ, k = z ˆβ, and (as always) e Summarizing: Equations 5.86, 5.87, and 5.88 all ˆx + y ˆy + z ˆz represent the same wave. They only appear to be r=x (5.91)

5.4. PROPAGATION OF A UNIFORM PLANE WAVE IN AN ARBITRARY DIRECTION Thus, k · r = βz, as expected. In ray-fixed coordinates, a wave can be represented by one – and only one – expression, which is the same expression regardless of the orientation of the “global” coordinate system. Moreover, only two basis ˆ and e ˆ) must be defined. Should directions (namely, k ˆ×e ˆ ˆ or e ˆ×k a third coordinate be required, either k may be selected as the additional basis direction. Note that the first choice has the possibly-useful feature that is the reference polarization of the e magnetic field intensity H. A general procedure for recasting the ray-fixed representation into a “coordinate-bound” representation is as follows. First, we represent k in the new fixed coordinate system; e.g.: ˆ + ky y ˆ + kz z ˆ k = kx x

(5.92)

where ˆ·x ˆ kx , β k ˆ·y ˆ ky , β k

(5.93) (5.94)

ˆ·z ˆ kz , β k

(5.95)

k · r = kx x + ky y + kz z

(5.96)

Then, With this expression in hand, Equation 5.89 may be rewritten as: e =e ˆE0 e−jkx x e−jky y e−jkz z E

(5.97)

ˆ = (ˆ ˆ) x ˆ + (ˆ ˆ) y ˆ + (ˆ ˆ) z ˆ e e·x e·y e·z

(5.98)

ˆ into its If desired, one can similarly decompose e Cartesian components as follows:

Thus, we see that the ray-fixed representation of Equation 5.89 accommodates all possible combinations of direction of propagation and reference polarization. Example 5.6. Plane wave propagating away from the z-axis. A uniform plane wave exhibiting a reference ˆ propagates away from the polarization of z z-axis. Develop representations of this wave in ray-fixed and global Cartesian coordinates.

69

Solution. As always, the ray-fixed representation is given by Equation 5.89. Since ˆ, e ˆ=z ˆ. the reference polarization is z Propagating away from the z-axis means ˆ=x ˆ cos φ + y ˆ sin φ k

(5.99)

where φ indicates the specific direction of propagation. For example, φ = 0 yields ˆ = +ˆ ˆ = +ˆ k x, φ = π/2 yields k y, and so on. Therefore, ˆ = β (ˆ ˆ sin φ) k , βk x cos φ + y

(5.100)

For completeness, note that the following factor appears in the phase-determining exponent in Equation 5.89: k · r = β (x cos φ + y sin φ)

(5.101)

In this case, we see kx = β cos φ, ky = β sin φ, and kz = 0. Thus, the wave may be expressed in Cartesian coordinates as follows: e =z ˆE0 e−jβx cos φ e−jβy sin φ E

(5.102)

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

70

5.5 Decomposition of a Wave into TE and TM Components [m0166]

A broad range of problems in electromagnetics involve scattering of a plane wave by a planar boundary between dissimilar media. Section 5.1 (“Plane Waves at Normal Incidence on a Planar Boundary Between Lossless Media”) addressed the special case in which the wave arrives in a direction which is perpendicular to the boundary (i.e., “normal incidence”). Analysis of the normal incidence case is simplified by the fact that the directions of field vectors associated with the reflected and transmitted are the same (except possibly with a sign change) as those of the incident wave. For the more general case in which the incident wave is obliquely incident (i.e., not necessarily normally-incident), the directions of the field vectors will generally be different. This added complexity is easily handled if we take the effort to represent the incident wave as the sum of two waves having particular polarizations. These polarizations are referred to as transverse electric (TE) and transverse magnetic (TM). This section describes these polarizations and the method for decomposition of a plane wave into TE and TM components. We will then be prepared to address the oblique incidence case in a later section. To begin, we define the ray-fixed coordinate system ˆ i is a unit vector shown in Figure 5.8. In this figure, k indicating the direction in which the incident wave ˆ is perpendicular to the propagates. The unit normal n boundary, and points into the region from which the wave is incident. We now make the following definition:

we see that:

ˆi × n ˆ k ˆ⊥ , e ˆi ˆ k × n

(5.103)

ˆ⊥ is a unit vector which is Defined in this manner, e ˆ i , and so may serve as a basis perpendicular to both k vector of a coordinate system which is attached to the incident ray. The remaining basis vector for this ray-fixed coordinate system is chosen as follows: ˆi ˆik , e ˆ⊥ × k e

(5.104)

ˆik is a unit vector which is Defined in this manner, e ˆ i , and parallel to the ˆ⊥ and k perpendicular to both e plane of incidence. Let us now examine an incident uniform plane wave in the new, ray-fixed coordinate system. We begin with the following phasor representation of the electric field intensity:

ˆi

i ei = e ˆi E0i e−jk ·r E

(5.105)

where e is a unit vector indicating the reference polarization, E0i is a complex-valued scalar, and r is a e i is vector indicating the position at which E i ˆ in the ray-fixed evaluated. We may express e coordinate system of Figure 5.8 as follows:  ˆi · e ˆ⊥ e ˆ⊥ ˆi = e e   ˆi · e ˆik e ˆik + e   ˆi k ˆi ˆi · k + e (5.106)

The plane of incidence is the plane in which both the normal to the surface (ˆ n) and the direction of ˆ i ) lie. propagation (k

The TE-TM decomposition consists of finding the components of the electric and magnetic fields which are perpendicular (“transverse”) to the plane of incidence. Of the two possible directions that are perpendicular to the plane of incidence, we choose ˆ⊥ , defined as shown in Figure 5.8. From the figure, e

c C. Wang CC BY-SA 4.0

Figure 5.8: Coordinate system for TE-TM decompoˆ i and n ˆ lie in the plane of the page, sition. Since both k this is also the plane of incidence.

5.5. DECOMPOSITION OF A WAVE INTO TE AND TM COMPONENTS The electric field vector is always perpendicular to the ˆ i = 0. This leaves: ˆi · k direction of propagation, so e    ˆi = e ˆi · e ˆ⊥ e ˆ⊥ + e ˆi · e ˆik e ˆik e (5.107) Substituting this expression into Equation 5.105, we obtain: i i ei = e ˆ⊥ ETi E e−jk ·r + e ˆik ETi M e−jk ·r E

ETi M

,

E0i

i

ˆ · e

ˆik e

Note that the nomenclature “TE” and “TM” is commonly but not universally used. Sometimes “TE” is referred to as “perpendicular” polarization, indicated using the subscript “⊥” or “s” (short for senkrecht, German for “perpendicular”). Correspondingly, “TM” is sometimes referred to as “parallel” polarization, indicated using the subscript “k” or “p.”

(5.108)

where

ˆi · e ˆ⊥ ETi E , E0i e

71

(5.109) (5.110)

The first term in Equation 5.108 is the transverse e i , so-named because it is electric (TE) component of E the component which is perpendicular to the plane of incidence. The second term in Equation 5.108 is the e i . The term transverse magnetic (TM) component of E “TM” refers to the fact that the magnetic field associated with this component of the electric field is perpendicular to the plane of incidence. This is apparent since the magnetic field vector is perpendicular to both the direction of propagation and the electric field vector. Summarizing: The TE component is the component for which e i is perpendicular to the plane of incidence. E The TM component is the component for which e i is perpendicular to the plane of incidence; H e i is parallel to the i.e., the component for which E plane of incidence. Finally, observe that the total wave is the sum of its TE and TM components. Therefore, we may analyze the TE and TM components separately, and know that the result for the combined wave is simply the sum of the results for the TE and TM components. As stated at the beginning of this section, the utility of the TE-TM decomposition is that it simplifies the analysis of wave reflection. This is because the analysis of the TE and TM cases is relatively simple, whereas direct analysis of the scattering of arbitrarily-polarized waves is relatively difficult.

Also, note that the TE component and TM component are sometimes referred to as the TE mode and TM mode, respectively. While the terms “component” and “mode” are synonymous for the single plane wave scenarios considered in this section, the terms are not synonymous in general. For example, the wave inside a waveguide may consist of multiple unique TE modes which collectively comprise the TE component of the field, and similarly the wave inside a waveguide may consist of multiple unique TM modes which collectively comprise the TM component of the field. Finally, consider what happens when a plane wave is normally-incident upon the boundary; i.e., when ˆ i = −ˆ k n. In this case, Equation 5.103 indicates that ˆ⊥ = 0, so the TE-TM decomposition is undefined. e The situation is simply that both Ei and Hi are already both perpendicular to the boundary, and there is no single plane that can be uniquely identified as the plane of incidence. We refer to this case as transverse electromagnetic (TEM). A wave which is normally-incident on a planar surface is said to be transverse electromagnetic (TEM) with respect to that boundary. There is no unique TE-TM decomposition in this case. The fact that a unique TE-TM decomposition does not exist in the TEM case is of no consequence, because the TEM case is easily handled as a separate condition (see Section 5.1).

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

72

5.6 Plane Waves at Oblique Incidence on a Planar Boundary: TE Case

ˆ i is the unit vector indicating the direction of where k √ propagation and β1 = ω µ1 ǫ1 is the phase e i serves as the propagation constant in Region 1. E TE “stimulus” in this problem, so all other contributions to the total field will be expressed in terms of quantities appearing in Equation 5.111.

[m0167]

In this section, we consider the problem of reflection and transmission from a planar boundary between semi-infinite media for a transverse electric (TE) uniform plane wave. Before attempting this section, a review of Sections 5.1 (“Plane Waves at Normal Incidence on a Planar Boundary Between Lossless Media”) and 5.5 (“Decomposition of a Wave into TE and TM Components”) is recommended. Also, note that this section has much in common with Section 5.7 (“Plane Waves at Oblique Incidence on a Planar Boundary: TM Case”), although it is recommended to attempt the TE case first.

The presence of reflected and transmitted uniform plane waves is inferred from our experience with the normal incidence scenario (Section 5.1). There, as here, the symmetry of the problem indicates that the reflected and transmitted components of the electric field will have the same polarization as that of the incident electric field. This is because there is nothing present in the problem that could account for a change in polarization. Thus, the reflected and transmitted fields will also be TE. Therefore, we postulate the following expression for the reflected wave: r e r (r) = y ˆ Be−jk ·r E

(5.113)

The TE case is illustrated in Figure 5.9. The boundary where B is an unknown, possibly complex-valued between the two semi-infinite and lossless regions is constant to be determined and located at the z = 0 plane. The wave is incident from ˆ r β1 e i of this kr = k (5.114) Region 1. The electric field intensity E TE wave is given by indicates the direction of propagation, which is also i i −jki ·r e ˆE e E (r) = y (5.111) currently unknown. TE

TE

Similarly, we postulate the following expression for e i is In this expression, r is the position at which E TE the transmitted wave: evaluated, and t ˆ i β1 e t (r) = y ki = k (5.112) ˆ Ce−jk ·r E (5.115) where C is an unknown, possibly complex-valued constant to be determined; ˆ t β2 kt = k

(5.116)

ˆ t is the unit vector indicating the direction of where k √ propagation; and β2 = ω µ2 ǫ2 is the phase propagation constant in Region 2. At this point, the unknowns in this problem are the ˆ r and constants B and C, as well as the directions k t ˆ k . We may establish a relationship between ETi E , B, and C by application of boundary conditions at c C. Wang CC BY-SA 4.0

z = 0. First, recall that the tangential component of the total electric field intensity must be continuous Figure 5.9: A TE uniform plane wave obliquely inci- across material boundaries. To apply this boundary e 1 and E e 2 to be the total dent on the planar boundary between two semi-infinite condition, let us define E material regions. electric fields in Regions 1 and 2, respectively. The

5.6. PLANE WAVES AT OBLIQUE INCIDENCE ON A PLANAR BOUNDARY: TE CASE

73

total field in Region 1 is the sum of incident and reflected fields, so e 1 (r) = E e i (r) + E e r (r) E TE

(5.117)

The total field in Region 2 is simply e 2 (r) = E e t (r) E

(5.118)

Next, note that all electric field components are already tangent to the boundary. Thus, continuity of the tangential component of the electric field across e 1 (0) = E e 2 (0), and therefore the boundary requires E e i (r0 ) + E e r (r0 ) = E e t (r0 ) E TE

(5.119)

ˆx + y ˆ y since z = 0 on the where r = r0 , x boundary. Now employing Equations 5.111, 5.113, and 5.115, we obtain: ˆ ETi E e−jk y

i

·r0

ˆ Be−jk +y

r

·r0

ˆ Ce−jk =y

t

c C. Wang CC BY-SA 4.0

Figure 5.10: Incident, reflected, and transmitted magnetic field components associated with the TE electric field components shown in Figure 5.9.

·r0

(5.120) obtained by applying the appropriate boundary conditions to the magnetic field. The magnetic field associated with each of the electric field components Dropping the vector (ˆ y) since it is the same in each is identified in Figure 5.10. Note the orientations of term, we obtain: the magnetic field vectors may be confirmed using the i r t ETi E e−jk ·r0 + Be−jk ·r0 = Ce−jk ·r0 (5.121) plane wave relationships: Specifically, the cross product of the electric and magnetic fields should point in the direction of propagation. Expressions for For Equation 5.121 to be true at every point r0 on the each of the magnetic field components is determined boundary, it must be true that formally below. i r t k · r0 = k · r0 = k · r0 (5.122) From the plane wave relationships, we determine that the incident magnetic field intensity is Essentially, we are requiring the phases of each field in Regions 1 and 2 to be matched at every point along ˆi × E ei e i (r) = 1 k the boundary. Any other choice will result in a H (5.124) TE η 1 violation of boundary conditions at some point along p the boundary. This phase matching criterion will where η1 = µ1 /ǫ1 is the wave impedance in determine the directions of propagation of the Region 1. To make progress requires that we express reflected and transmitted fields, which we shall do ˆ i in the global fixed coordinate system. Here it is: k later. ˆi = x ˆ sin ψ i + z ˆ cos ψ i k (5.125) First, let us use the phase matching criterion to complete the solution for the coefficients B and C. Thus: Enforcing Equation 5.122, we observe that Equation 5.121 reduces to:  ETi E −jki ·r e i (r) = z ˆ sin ψ i − x ˆ cos ψ i H e (5.126) η1 (5.123) ETi E + B = C Similarly, we determine that the reflected magnetic field has the form: A second equation is needed since we currently have only one equation (Equation 5.123) and two unknowns (B and C). The second equation is

ˆr × E e r (r) = 1 k er H η1

(5.127)

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

74

Equations 5.123 and 5.137 comprise a linear system of equations with unknowns B and C. This system of (5.128) equations is easily solved for B as follows. First, use Thus: Equation 5.123 to eliminate C in Equation 5.137. The result is: r B e r (r) = (ˆ ˆ cos ψ r ) e−jk ·r (5.129) H z sin ψ r + x η1  ETi E − cos ψ i The transmitted magnetic field has the form: η1 1 B ˆt × E e t (r) = k et H (5.130) + (cos ψ r ) η2 η1  ETi E + B In the global coordinate system: = − cos ψ t (5.138) η2 ˆt = x ˆ sin ψ t + z ˆ cos ψ t k (5.131) In the global coordinate system: ˆr = x ˆ sin ψ r − z ˆ cos ψ r k

Solving this equation for B, we obtain:

Thus:  C −jkt ·r e t (r) = z ˆ sin ψ t − x ˆ cos ψ t H e η2

(5.132)

B=

η2 cos ψ i − η1 cos ψ t i E η2 cos ψ r + η1 cos ψ t T E

(5.139)

The total magnetic field in Region 1 is the sum of incident and reflected fields, so e 1 (r) = H e i (r) + H e r (r) H

We can express this result as a reflection coefficient as follows: (5.133) B = ΓT E ETi E (5.140)

The magnetic field in Region 2 is simply e 2 (r) = H e t (r) H

where (5.134)

Since there is no current on the boundary, the tangential component of the total magnetic field intensity must be continuous across the boundary. Expressed in terms of the quantities already established, this boundary condition requires: e i (r0 ) + x e r (r0 ) = x e t (r0 ) ˆ·H ˆ·H ˆ·H x

 ETi E − cos ψ i η1 B + (cos ψ r ) η1  t C = − cos ψ η2

(5.141)

It is worth noting that Equation 5.141 becomes the reflection coefficient for normal (TEM) incidence when ψ i = ψ r = ψ t = 0, as expected.

C = (1 + ΓT E ) ETi E

(5.142)

Let us now summarize the solution. Given the TE electric field intensity expressed in Equation 5.111, we find:

− cos ψ i

Now employing the phase matching condition expressed in Equation 5.122, we find:

η2 cos ψ i − η1 cos ψ t η2 cos ψ r + η1 cos ψ t

(5.135) Returning to Equation 5.123, we now find

where “ˆ x·” selects the component of the magnetic field that is tangent to the boundary. Evaluating this expression, we obtain:  ETi E −jki ·r0 e η1 r B + (cos ψ r ) e−jk ·r0 η1  C −jkt ·r0 = − cos ψ t e η2

ΓT E ,

(5.136)

r e r (r) = y ˆ ΓT E ETi E e−jk ·r E

t e t (r) = y ˆ (1 + ΓT E ) ETi E e−jk ·r E

(5.143) (5.144)

This solution is complete except that we have not yet ˆ r , which is now completely determined determined k r ˆ t , which is now by ψ via Equation 5.128, and k t completely determined by ψ via Equation 5.131. In other words, we have not yet determined the directions of propagation ψ r for the reflected wave t r i (5.137) and ψ for the transmitted wave. However, ψ and ψ can be found using Equation 5.122. Here we shall

5.6. PLANE WAVES AT OBLIQUE INCIDENCE ON A PLANAR BOUNDARY: TE CASE

75

simply state the result, and in Section 5.8 we shall perform this part of the derivation in detail and with greater attention to the implications. One finds: ψr = ψi and t

ψ = arcsin



β1 sin ψ i β2

(5.145) 

(5.146)

Equation 5.145 is the unsurprising result that angle of reflection equals angle of incidence. Equation 5.146 – addressing angle of transmission – is a bit more intriguing. Astute readers may notice that there is something fishy about this equation: It seems possible for the argument of arcsin to be greater than one. This oddity is addressed in Section 5.8.

c C. Wang CC BY-SA 4.0

Figure 5.11: A TE uniform plane wave incident from air to glass.

Finally, note that Equation 5.145 allows us to eliminate ψ r from Equation 5.141, yielding: ΓT E =

η2 cos ψ i − η1 cos ψ t η2 cos ψ i + η1 cos ψ t

η0 ∼ η2 ≈ √ = 260.0 Ω (glass) 2.1 ψ i = 30◦

(5.147)

(5.150)

Note √ ω µ0 ǫ0 ∼ β1 ≈ √ = 0.690 β2 ω µ0 · 2.1ǫ0

Thus, we obtain what is perhaps the most important finding of this section: The electric field reflection coefficient for oblique TE incidence, ΓT E , is given by Equation 5.147. The following example demonstrates the utility of this result.



β1 sin ψ i β2



∼ = 20.2◦

Figure 5.11 illustrates a TE plane wave incident from air onto the planar boundary with glass. The glass exhibits relative permittivity of 2.1. Determine the power reflected and transmitted relative to power incident on the boundary. Solution. The power reflected relative to power 2 incident is |ΓT E | whereas the power transmitted relative to power incident is 2 1 − |ΓT E | . ΓT E may be calculated using Equation 5.147. Calculating the quantities that enter into this expression: (5.148)

(5.152)

Now substituting these values into Equation 5.147, we obtain ΓT E ∼ = −0.2220

Example 5.7. Power transmission at an air-to-glass interface (TE case).

(5.151)

so ψ t = arcsin

η1 ≈ η0 ∼ = 376.7 Ω (air)

(5.149)

(5.153)

Subsequently, the fraction of power reflected 2 relative to power incident is |ΓT E | ∼ = 0.049; 2 ∼ i.e., about 4.9%. 1 − |ΓT E | = 95.1% of the power is transmitted into the glass.

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

76

5.7 Plane Waves at Oblique Incidence on a Planar Boundary: TM Case

evaluated. Also, ˆ i β1 ki = k

[m0164]

In this section, we consider the problem of reflection and transmission from a planar boundary between semi-infinite media for a transverse magnetic (TM) uniform plane wave. Before attempting this section, a review of Sections 5.1 (“Plane Waves at Normal Incidence on a Planar Boundary Between Lossless Media”) and 5.5 (“Decomposition of a Wave into TE and TM Components”) is recommended. Also, note that this section has much in common with Section 5.6 (“Plane Waves at Oblique Incidence on a Planar Boundary: TE Case”), and it is recommended to attempt the TE case first.

(5.155)

ˆ i is the unit vector indicating the direction of where k √ propagation and β1 = ω µ1 ǫ1 is the phase e i serves as the propagation constant in Region 1. H TM “stimulus” in this problem, and all other contributions to the total field may be expressed in terms of parameters associated with Equation 5.154. The presence of reflected and transmitted uniform plane waves is inferred from our experience with the normal incidence scenario (Section 5.1). There, as here, the symmetry of the problem indicates that the reflected and transmitted components of the magnetic field will have the same polarization as that of the incident electric field. This is because there is nothing present in the problem that could account for a change in polarization. Thus, the reflected and transmitted fields will also be TM. So we postulate the following expression for the reflected wave:

In this section, we consider the scenario illustrated in r Figure 5.12. The boundary between the two e r (r) = −ˆ H yBe−jk ·r (5.156) semi-infinite and lossless regions is located at the z = 0 plane. The wave is incident from Region 1. The where B is an unknown, possibly complex-valued e i of this wave is given by constant to be determined and magnetic field intensity H TM i e i (r) = y ˆ HTi M e−jk ·r H TM

(5.154)

e i is In this expression, r is the position at which H TM

ˆ r β1 kr = k

(5.157)

indicates the direction of propagation. The reader may wonder why we have chosen −ˆ y, as e r. opposed to +ˆ y, as the reference polarization for H In fact, either +ˆ y or −ˆ y could be used. However, the choice is important because the form of the results we obtain in this section – specifically, the reflection coefficient – will be determined with respect to this specific convention, and will be incorrect with respect to the opposite convention. We choose −ˆ y because it has a particular advantage which we shall point out at the end of this section. Continuing, we postulate the following expression for the transmitted wave:

c C. Wang CC BY-SA 4.0

t e t (r) = y ˆ Ce−jk ·r H

(5.158)

where C is an unknown, possibly complex-valued Figure 5.12: A TM uniform plane wave obliquely constant to be determined and incident on the planar boundary between two semiˆ t β2 infinite material regions. kt = k (5.159)

5.7. PLANE WAVES AT OBLIQUE INCIDENCE ON A PLANAR BOUNDARY: TM CASE ˆ t is the unit vector indicating the direction of where k √ propagation and β2 = ω µ2 ǫ2 is the phase propagation constant in Region 2. At this point, the unknowns in this problem are the constants B and C, as well as the unknown directions ˆ r and k ˆ t . We may establish a relationship between k i HT M , B, and C by application of boundary conditions at z = 0. First, we presume no impressed current at the boundary. Thus, the tangential component of the total magnetic field intensity must be continuous across the boundary. To apply this e 1 and H e 2 to be the boundary condition, let us define H total magnetic fields in Regions 1 and 2, respectively. The total field in Region 1 is the sum of incident and reflected fields, so

77

the directions of propagation of the reflected and transmitted fields, which we shall do later. Our priority for now shall be to solve for the coefficients B and C. Enforcing Equation 5.165, we observe that Equation 5.164 reduces to: HTi M − B = C

(5.166)

A second equation is needed since we currently have only one equation (Equation 5.166) and two unknowns (B and C). The second equation is obtained by applying the appropriate boundary conditions to the electric field. The electric field associated with each of the magnetic field i r e e e H1 (r) = HT M (r) + H (r) (5.160) components is identified in Figure 5.12. Note the orientations of the electric field vectors may be confirmed using the plane wave relationships: The field in Region 2 is simply Specifically, the cross product of the electric and e 2 (r) = H e t (r) H (5.161) magnetic fields should point in the direction of propagation. Expressions for each of the electric field Also, we note that all magnetic field components are components is determined formally below. already tangent to the boundary. Thus, continuity of the tangential component of the magnetic field across From the plane wave relationships, we determine that e 1 (r0 ) = H e 2 (r0 ), where the boundary requires H the incident electric field intensity is ˆx + y ˆ y since z = 0 on the boundary. r0 , x ˆi × H e i (r) = −η1 k ei E (5.167) Therefore, TM p e i (r0 ) + H e r (r0 ) = H e t (r0 ) H (5.162) where η1 = µ1 /ǫ1 is the wave impedance in TM Region 1. To make progress requires that we express ˆ i in the global fixed coordinate system. Here it is: Now employing Equations 5.154, 5.156, and 5.158, k we obtain: ˆi = x ˆ sin ψ i + z ˆ cos ψ i k (5.168) i r t ˆ HTi M e−jk ·r0 − y ˆ Be−jk ·r0 = y ˆ Ce−jk ·r0 y (5.163) Thus: Dropping the vector (ˆ y) since it is the same in each  i e i (r) = x ˆ cos ψ i − z ˆ sin ψ i η1 HTi M e−jk ·r E term, we obtain: (5.169) i r t HTi M e−jk ·r0 − Be−jk ·r0 = Ce−jk ·r0 (5.164) Similarly we determine that the reflected electric field has the form: For this to be true at every point r0 on the boundary, it ˆr × H e r (r) = −η1 k er must be true that E (5.170) k i · r0 = k r · r0 = k t · r0

(5.165) In the global coordinate system:

Essentially, we are requiring the phases of each field in Regions 1 and 2 to be matched at every point along the boundary. Any other choice will result in a violation of boundary conditions at some point along the boundary. This expression allows us to solve for

ˆr = x ˆ sin ψ r − z ˆ cos ψ r k

(5.171)

Thus: r e r (r) = (ˆ ˆ sin ψ r ) η1 Be−jk ·r (5.172) E x cos ψ r + z

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

78

Solving this equation for B, we obtain:

The transmitted magnetic field has the form: ˆt × H e t (r) = −η2 k et E

(5.173)

ˆt = x ˆ sin ψ t + z ˆ cos ψ t k

(5.174)

In the global coordinate system:

B=

−η1 cos ψ i + η2 cos ψ t i H +η1 cos ψ r + η2 cos ψ t T M

We can express this result as follows: B = ΓT M HTi M

Thus:

(5.183)

 t e t (r) = x ˆ cos ψ t − z ˆ sin ψ t η2 Ce−jk ·r (5.175) where we have made the definition E −η1 cos ψ i + η2 cos ψ t The total electric field in Region 1 is the sum of ΓT M , +η1 cos ψ r + η2 cos ψ t incident and reflected fields, so e 1 (r) = E e i (r) + E e r (r) E

(5.182)

(5.184)

(5.176)

We are now able to express the complete solution in terms of the electric field intensity. First we make the The electric field in Region 2 is simply substitution ETi M , η1 HTi M in Equation 5.169, e 2 (r) = E e t (r) E (5.177) yielding:  i The tangential component of the total electric field e i (r) = x ˆ cos ψ i − z ˆ sin ψ i ETi M e−jk ·r (5.185) E intensity must be continuous across the boundary. Expressed in terms of the quantities already The factor η1 B in Equation 5.172 becomes established, this boundary condition requires: ΓT M ETi M , so we obtain: e i (r0 ) + x e r (r0 ) = x e t (r0 ) ˆ·E ˆ·E ˆ·E x

(5.178)

where “ˆ x·” selects the component of the electric field that is tangent to the boundary. Evaluating this expression, we obtain:  i + cos ψ i η1 HTi M e−jk ·r0 r

+ (cos ψ r ) η1 Be−jk ·r0  t = + cos ψ t η2 Ce−jk ·r0

(5.179)

Now employing the “phase matching” condition expressed in Equation 5.165, we find:  + cos ψ i η1 HTi M + (cos ψ r ) η1 B  = + cos ψ t η2 C

e r (r) = (ˆ ˆ sin ψ r ) E x cos ψ r + z · ΓT M ETi M e−jk

r

·r

(5.186)

Thus, we see ΓT M is the reflection coefficient for the electric field intensity. Returning to Equation 5.166, we now find C = HTi M − B

= HTi M − ΓT M HTi M = (1 − ΓT M ) HTi M

= (1 − ΓT M ) ETi M /η1

(5.187)

Subsequently, Equation 5.175 becomes  (5.180) e t (r) = x ˆ cos ψ t − z ˆ sin ψ t E t η2 · (1 − ΓT M ) ETi M e−jk ·r (5.188) Equations 5.166 and 5.180 comprise a linear system η1 of equations with unknowns B and C. This system of equations is easily solved for B as follows. First, use Equation 5.166 to eliminate C in Equation 5.180. The This solutionˆ ris complete except that we have not yet determined k , which is now completely determined result is: ˆ t , which is now by ψ r via Equation 5.171, and k  t i i completely determined by ψ via Equation 5.174. In + cos ψ η1 HT M r other words, we have not yet determined the + (cos ψ ) η1 B directions of propagation ψ r for the reflected wave   (5.181) and ψ t for the transmitted wave. However, ψ r and ψ i = + cos ψ t η2 HTi M − B

5.7. PLANE WAVES AT OBLIQUE INCIDENCE ON A PLANAR BOUNDARY: TM CASE can be found using Equation 5.165. Here we shall simply state the result, and in Section 5.8 we shall perform this part of the derivation in detail and with greater attention to the implications. One finds: ψr = ψi

(5.189)

i.e., angle of reflection equals angle of incidence. Also,   β1 ψ t = arcsin (5.190) sin ψ i β2 Astute readers may notice that there is something fishy about Equation 5.190. Namely, it seems possible for the argument of arcsin to be greater than one. This oddity is addressed in Section 5.8. Now let us return to the following question, raised near the beginning of this section: Why choose −ˆ y, as opposed to +ˆ y, as the reference polarization for Hr , as shown in Figure 5.12? To answer this question, first note that ΓT M (Equation 5.184) becomes the reflection coefficient for normal (TEM) incidence when ψ i = ψ t = 0. If we had chosen +ˆ y as the reference polarization for Hr , we would have instead obtained an expression for ΓT M that has the opposite sign for TEM incidence.3 There is nothing wrong with this answer, but it is awkward to have different values of the reflection coefficient for the same physical scenario. By choosing −ˆ y, the reflection coefficient for the oblique incidence case computed for ψ i = 0 converges to the reflection coefficient previously computed for the normal-incidence case. It is important to be aware of this issue, as one occasionally encounters work in which the opposite (“+ˆ y”) reference polarization has been employed. Finally, note that Equation 5.189 allows us to eliminate ψ r from Equation 5.184, yielding: i

ΓT M =

−η1 cos ψ + η2 cos ψ +η1 cos ψ i + η2 cos ψ t

The following example demonstrates the utility of this result. Example 5.8. Power transmission at an air-to-glass interface (TM case). Figure 5.13 illustrates a TM plane wave incident from air onto the planar boundary with a glass region. The glass exhibits relative permittivity of 2.1. Determine the power reflected and transmitted relative to power incident on the boundary. Solution. The power reflected relative to power 2 incident is |ΓT M | whereas the power transmitted relative to power incident is 2 1 − |ΓT M | . ΓT M may be calculated using Equation 5.191. Calculating the quantities that enter into this expression: η1 ≈ η0 ∼ = 376.7 Ω (air)

(5.192)

η0 ∼ η2 ≈ √ = 260.0 Ω (glass) 2.1

(5.193)

ψ i = 30◦

(5.194)

Note √ ω µ0 ǫ0 ∼ β1 ≈ √ = 0.690 β2 ω µ0 · 2.1ǫ0

(5.191)

Thus, we obtain what is perhaps the most important finding of this section: The electric field reflection coefficient for oblique TM incidence, ΓT M , is given by Equation 5.191. 3 Obtaining this result is an excellent way for the student to confirm their understanding of the derivation presented in this section.

(5.195)

so t

ψ = arcsin



β1 sin ψ i β2



∼ = 20.2◦

(5.196)

Now substituting these values into Equation 5.191, we obtain ΓT M ∼ = −0.1442

t

79

(5.197)

(Did you get an answer closer to −0.1323? If so, you probably did not use sufficient precision to represent intermediate results. This is a good example of a problem in which three significant figures for results that are used in subsequent calculations is not sufficient.) The fraction of power reflected relative to power incident is now determined to be

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

80

5.8 Angles of Reflection and Refraction [m0168]

Consider the situation shown in Figure 5.14: A uniform plane wave obliquely incident on the planar boundary between two semi-infinite material regions. Let a point on the boundary be represented as the position vector c C. Wang CC BY-SA 4.0

ˆx + y ˆy r0 = x

(5.198)

Figure 5.13: A TM uniform plane wave incident from In both Sections 5.6 (“Plane Waves at Oblique air to glass. Incidence on a Planar Boundary: TE Case”) and 5.7 (“Plane Waves at Oblique Incidence on a Planar 2 Boundary: TM Case”), it is found that |ΓT M | ∼ 0.021; i.e., about 2.1%. = 2 ∼ 1 − |ΓT M | = 97.9% of the power is transmitted k i · r0 = k r · r0 = k t · r0 (5.199) into the glass. In this expression, Note that the result obtained in the preceding example is different from the result for a TE wave incident from the same direction (Example 5.7). In other words: The fraction of power reflected and transmitted from the planar boundary between dissimilar media depends on the polarization of the incident wave relative to the boundary, as well as the angle of incidence.

ˆi k i = β1 k ˆr k r = β1 k

(5.200)

ˆt k t = β2 k

(5.202)

(5.201)

ˆi, k ˆ r , and k ˆ t are unit vectors in the direction where k of incidence, reflection, and transmission,

c C. Wang CC BY-SA 4.0

Figure 5.14: A uniform plane wave obliquely incident on the planar boundary between two semi-infinite material regions.

5.8. ANGLES OF REFLECTION AND REFRACTION respectively; and β1 and β2 are the phase propagation constants in Region 1 (from which the wave is incident) and Region 2, respectively. Equation 5.199 is essentially a boundary condition that enforces continuity of the phase of the electric and magnetic fields across the boundary, and is sometimes referred to as the “phase matching” requirement. Since the same requirement emerges independently in the TE and TM cases, and since any plane wave may be decomposed into TE and TM components, the requirement must apply to any incident plane wave regardless of polarization. Equation 5.199 is the key to finding the direction of reflection ψ r and direction of transmission ψ t . First, observe:

81

Thus, we see that ψ t does not depend on frequency, except to the extent that the constitutive parameters might. We may also express this relationship in terms of the relative values of constitutive parameters; i.e., µ1 = µr1 µ0 , ǫ1 = ǫr1 ǫ0 , µ2 = µr2 µ0 , and ǫ2 = ǫr2 ǫ0 . In terms of the relative parameters: √ √ µr1 ǫr1 sin ψ i = µr2 ǫr2 sin ψ t

(5.211)

This is known as Snell’s law or the law of refraction. Refraction is simply transmission with the result that the direction of propagation is changed. Snell’s law (Equation 5.211) determines the angle of refraction (transmission).

The associated formula for ψ t explicitly is:  r µr1 ǫr1 (5.204) (5.212) sin ψ i ψ t = arcsin µr2 ǫr2 (5.205) For the common special case of non-magnetic media, Therefore, we may express Equation 5.199 in the one assumes µr1 = µr2 = 1. In this case, Snell’s law following form simplifies to:  √ √ ǫr1 sin ψ i = ǫr2 sin ψ t (5.213) ˆ sin ψ i + z ˆ cos ψ i · (ˆ ˆ y) β1 x xx + y ˆi = x ˆ sin ψ i + z ˆ cos ψ i k ˆr = x ˆ sin ψ r − z ˆ cos ψ r k ˆt = x ˆ sin ψ t + z ˆ cos ψ t k

ˆ cos ψ r ) · (ˆ ˆ y) = β1 (ˆ x sin ψ r − z xx + y  t t ˆ sin ψ + z ˆ cos ψ · (ˆ ˆ y) = β2 x xx + y

which reduces to

β1 sin ψ i = β1 sin ψ r = β2 sin ψ t

(5.203)

In optics, it is common to express permittivities in √ (5.206) terms of indices of refraction; e.g., n1 , ǫr1 and √ n2 , ǫr2 . Thus, Snell’s law in optics is often expressed as: (5.207) n sin ψ i = n sin ψ t (5.214) 1

2

Examining the first and second terms of Equation 5.207, and noting that ψ i and ψ r are both limited to the range −π/2 to +π/2, we find that: ψr = ψi In plain English:

When both media are non-magnetic, Equation 5.212 simplifies to r  ǫr1 i t (5.208) ψ = arcsin sin ψ (5.215) ǫr2

When ǫr2 > ǫr1 , we observe that ψ t < ψ i . In other words, the transmitted wave travels in a direction that is closer to the surface normal than the angle of incidence. This scenario is demonstrated in the Examining the first and third terms of Equation 5.207, following example. we find that Angle of reflection equals angle of incidence.

β1 sin ψ i = β2 sin ψ t (5.209) √ √ Since β1 = ω µ1 ǫ1 and β2 = ω µ2 ǫ2 , Equation 5.209 expressed explicitly in terms of the constitutive parameters is: √ √ µ1 ǫ1 sin ψ i = µ2 ǫ2 sin ψ t (5.210)

Example 5.9. Refraction of light at an air-glass boundary. Figure 5.15 shows a narrow beam of light that is incident from air to glass at an angle ψ i = 60◦ .

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

82

c G. Saini CC BY-SA 4.0

c Z. S´andor CC BY-SA 3.0

Figure 5.15: Angles of reflection and refraction for a light wave incident from air onto glass. As expected, the angle of reflection ψ r is observed to be equal to ψ i . The angle of refraction ψ t is observed to be 35◦ . What is the relative permittivity of the glass? Solution. Since the permittivity of glass is greater than that of air, we observe the expected result ψ t < ψ i . Since glass is non-magnetic, the expected relationship between these angles is given by Equation 5.213. Solving that equation for the relative permittivity of the glass, we obtain: 2  sin ψ i ∼ (5.216) ǫr2 = ǫr1 = 2.28 sin ψ t

When a wave travels in the reverse direction – i.e., from a non-magnetic medium of higher permittivity to a medium of lower permittivity – one finds ψ t > ψ r . In other words, the refraction is away from the surface normal. Figure 5.16 shows an example from common experience. In non-magnetic media, when ǫr1 < ǫr2 , ψt < ψi (refraction toward the surface normal). When ǫr1 > ǫr2 , ψt > ψi (refraction away from the surface normal).

Figure 5.16: Refraction accounts for the apparent displacement of an underwater object from the perspective of an observer above the water. Under certain conditions, the ǫr2 < ǫr1 case leads to the following surprising observation: When calculating ψ tp using, for example, Equation 5.212, one finds that ǫr1 /ǫr2 sin ψ i can be greater than 1. Since the sin function yields values between −1 and +1, the result of the arcsin function is undefined. This odd situation is addressed in Section 5.11. For now, we will simply note that this condition leads to the phenomenon of total internal reflection. For now, all we can say is that when ǫr2 < ǫr1 , ψ t is able to reach π/2 radians, which corresponds to propagation parallel to the boundary. Beyond that threshold, we must account for the unique physical considerations associated with total internal reflection. We conclude this section with a description of the common waveguiding device known as the prism, shown in Figure 5.17. This particular device uses refraction to change the direction of light waves (similar devices can be used to manipulate radio waves as well). Many readers are familiar with the use of prisms to separate white light into its constituent colors (frequencies), as shown in Figure 5.18. The separation of colors is due to frequency dependence of the material comprising the prism. Specifically, the permittivity of the material is a function of frequency, and therefore the angle of refraction is a function of frequency. Thus, each frequency is refracted by a different amount. Conversely, a prism comprised of a material whose permittivity exhibits negligible variation with frequency will not separate incident

5.9. TE REFLECTION IN NON-MAGNETIC MEDIA

83

5.9 TE Reflection in Non-magnetic Media [m0171]

Figure 5.19 shows a TE uniform plane wave incident on the planar boundary between two semi-infinite material regions. In this case, the reflection coefficient is given by: ΓT E =

c D-Kuru CC BY-SA 3.0

Figure 5.17: A typical triangular prism. white light into its constituent colors since each color will be refracted by the same amount. Additional Reading: • “Prism” on Wikipedia. • “Refraction” on Wikipedia. • “Refractive index” on Wikipedia.

η2 cos ψ i − η1 cos ψ t η2 cos ψ i + η1 cos ψ t

(5.217)

where ψ i and ψ t are the angles of incidence and transmission (refraction), respectively; and η1 and η2 are the wave impedances in Regions 1 and 2, respectively. Many materials of practical interest are non-magnetic; that is, they have permeability that is not significantly different from the permeability of free space. In this section, we consider the behavior of the reflection coefficient for this class of materials. To begin, recall the general form of Snell’s law: sin ψ t =

β1 sin ψ i β2

(5.218)

In non-magnetic media, the permeabilities µ1 and µ2 are assumed equal to µ0 . Thus: r √ ω µ1 ǫ1 β1 ǫ1 = √ = (5.219) β2 ω µ2 ǫ2 ǫ2

• “Snell’s law” on Wikipedia. • “Total internal reflection” on Wikipedia.

c C. Wang CC BY-SA 4.0

c Suidroot CC BY-SA 4.0

Figure 5.18: A color-separating prism.

Figure 5.19: A TE uniform plane wave obliquely incident on the planar boundary between two semi-infinite regions of lossless non-magnetic material.

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

Since permittivity ǫ can be expressed as ǫ0 times the relative permittivity ǫr , we may reduce further to: r β1 ǫr1 = (5.220) β2 ǫr2 Now Equation 5.218 reduces to: r ǫr1 t sin ψ = sin ψ i ǫr2

(5.221)

1

0.5

Reflection Coefficient

84

0

2

10

-0.5

100

Next, note that for any value ψ, one may write cosine in terms of sine as follows: q cos ψ = 1 − sin2 ψ (5.222)

-1 0

10

20

30

40

50

60

70

80

90

angle of incidence [deg]

Therefore,

t

cos ψ =

r

1−

ǫr1 sin2 ψ i ǫr2

Also we note that in non-magnetic media r r µ1 µ0 η0 = =√ η1 = ǫ1 ǫr1 ǫ0 ǫr1 r r µ2 µ0 η0 η2 = = =√ ǫ2 ǫr2 ǫ0 ǫr2

(5.223)

(5.224) (5.225)

where η0 is the wave impedance in free space. Making substitutions into Equation 5.217, we obtain: √  √  η0 / ǫr2 cos ψ i − η0 / ǫr1 cos ψ t ΓT E = √  √  η0 / ǫr2 cos ψ i + η0 / ǫr1 cos ψ t (5.226) √ Multiplying numerator and denominator by ǫr2 /η0 , we obtain: p cos ψ i − ǫr2 /ǫr1 cos ψ t p ΓT E = (5.227) cos ψ i + ǫr2 /ǫr1 cos ψ t

Finally, by substituting Equation 5.223, we obtain:

ΓT E

q

ǫr2 /ǫr1 − sin2 ψ i q = cos ψ i + ǫr2 /ǫr1 − sin2 ψ i cos ψ i −

Figure 5.20: The reflection coefficient ΓT E as a function of angle of incidence ψ i for various media combinations. Curves are labeled with the ratio ǫr2 /ǫr1 . coefficient. First, we note that when ǫr1 = ǫr2 (i.e., same media on both sides of the boundary), ΓT E = 0 as expected. When ǫr1 > ǫr2 (e.g., wave traveling in glass toward air), we see that it is possible for ǫr2 /ǫr1 − sin2 ψ i to be negative, which makes ΓT E complex-valued. This results in total internal reflection, and is addressed elsewhere in another section. When ǫr1 < ǫr2 (e.g., wave traveling in air toward glass), we see that ǫr2 /ǫr1 − sin2 ψ i is always positive, so ΓT E is always real-valued. Let us continue with the ǫr1 < ǫr2 condition. Figure 5.20 shows ΓT E plotted for various combinations of media over all possible angles of incidence from 0 (normal incidence) to π/2 (grazing incidence). We observe: In non-magnetic media with ǫr1 < ǫr2 , ΓT E is real-valued, negative, and decreases to −1 as ψ i approaches grazing incidence.

(5.228)

This expression has the advantage that it is now entirely in terms of ψ i , with no need to first calculate ψt . Using Equation 5.228, we can see how different combinations of material affect the reflection

Also note that at any particular angle of incidence, ΓT E trends toward −1 as ǫr2 /ǫr1 increases. Thus, we observe that as ǫr2 /ǫr1 → ∞, the result is increasingly similar to the result we would obtain for a perfect conductor in Region 2.

5.10. TM REFLECTION IN NON-MAGNETIC MEDIA

5.10 TM Reflection in Non-magnetic Media [m0172]

85

Since permittivity ǫ can be expressed as ǫ0 times the relative permittivity ǫr , we may reduce further to: r β1 ǫr1 = (5.232) β2 ǫr2 Now Equation 5.230 reduces to: r ǫr1 t sin ψ = sin ψ i ǫr2

Figure 5.21 shows a TM uniform plane wave incident (5.233) on the planar boundary between two semi-infinite material regions. In this case, the reflection coefficient Next, note that for any value ψ, one may write cosine is given by: in terms of sine as follows: q −η1 cos ψ i + η2 cos ψ t cos ψ = 1 − sin2 ψ (5.234) ΓT M = (5.229) +η1 cos ψ i + η2 cos ψ t Therefore, r where ψ i and ψ t are the angles of incidence and ǫr1 t 1− cos ψ = sin2 ψ i (5.235) transmission (refraction), respectively; and η1 and η2 ǫr2 are the wave impedances in Regions 1 and 2, Also we note that in non-magnetic media respectively. Many materials of practical interest are r r non-magnetic; that is, they have permeability that is µ1 µ0 η0 η1 = = =√ (5.236) not significantly different from the permeability of ǫ1 ǫr1 ǫ0 ǫr1 r r free space. In this section, we consider the behavior µ2 µ0 η0 of the reflection coefficient for this class of materials. η2 = = =√ (5.237) ǫ2 ǫr2 ǫ0 ǫr2 To begin, recall the general form of Snell’s law: where η0 is the wave impedance in free space. Making substitutions into Equation 5.229, we obtain: β1 t i √  √  sin ψ = sin ψ (5.230) − η0 / ǫr1 cos ψ i + η0 / ǫr2 cos ψ t β2 ΓT M = √  √  + η0 / ǫr1 cos ψ i + η0 / ǫr2 cos ψ t In non-magnetic media, the permeabilities µ1 and µ2 (5.238) √ are assumed equal to µ0 . Thus: Multiplying numerator and denominator by ǫr2 /η0 , r √ we obtain: ω µ1 ǫ1 β1 ǫ1 p = √ = (5.231) − ǫr2 /ǫr1 cos ψ i + cos ψ t β2 ω µ2 ǫ2 ǫ2 (5.239) ΓT M = p + ǫr2 /ǫr1 cos ψ i + cos ψ t

Substituting Equation 5.235, we obtain: q p − ǫr2 /ǫr1 cos ψ i + 1 − (ǫr1 /ǫr2 ) sin2 ψ i q ΓT M = p + ǫr2 /ǫr1 cos ψ i + 1 − (ǫr1 /ǫr2 ) sin2 ψ i (5.240) This expression has the advantage that it is now entirely in terms of ψ i , with no need to first calculate ψt .

Finally, multiplying numerator and denominator by p ǫr2 /ǫr1 , we obtain: q c C. Wang CC BY-SA 4.0

− (ǫr2 /ǫr1 ) cos ψ i + ǫr2 /ǫr1 − sin2 ψ i q ΓT M = Figure 5.21: A transverse magnetic uniform plane + (ǫr2 /ǫr1 ) cos ψ i + ǫr2 /ǫr1 − sin2 ψ i wave obliquely incident on the planar boundary between two semi-infinite material regions. (5.241)

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

86 1

Reflection Coefficient

0.5

0

2

10

-0.5

100

-1 0

10

20

30

40

50

60

70

80

90

angle of incidence [deg]

c C. Wang CC BY-SA 4.0

Figure 5.23: Reflection of a plane wave with angle of i incidence ψ i equal to the polarizing angle ψB . Here Figure 5.22: The reflection coefficient ΓT M as a func- the media are non-magnetic with ǫ < ǫ . r1 r2 tion of angle of incidence ψ i for various media combinations, parameterized as ǫr2 /ǫr1 . ǫr2 /ǫr1 → ∞, the result for both the TE and TM components are increasingly similar to the result we Using Equation 5.241, we can easily see how would obtain for a perfect conductor in Region 2. different combinations of material affect the reflection Also note that when ǫr1 < ǫr2 , ΓT M changes sign coefficient. First, we note that when ǫr1 = ǫr2 (i.e., same media on both sides of the boundary), ΓT M = 0 from negative to positive as angle of incidence increases from 0 to π/2. This behavior is quite as expected. When ǫr1 > ǫr2 (e.g., wave traveling in different from that of the TE component, which is glass toward air), we see that it is possible for always negative for ǫr1 < ǫr2 . The angle of incidence ǫr2 /ǫr1 − sin2 ψ i to be negative, which makes ΓT M at which ΓT M = 0 is referred to as Brewster’s angle, complex-valued. This results in total internal i which we assign the symbol ψB . Thus: reflection, and is addressed in another section. When ǫr1 < ǫr2 (e.g., wave traveling in air toward glass), i we see that ǫr2 /ǫr1 − sin2 ψ i is always positive, so ψB , ψ i at which ΓT M = 0 (5.242) ΓT M is always real-valued. In the discussion that follows, here is the key point to Let us continue with the ǫr1 < ǫr2 condition. keep in mind: Figure 5.22 shows ΓT M plotted for various i combinations of media over all possible angles of Brewster’s angle ψB is the angle of incidence at incidence from 0 (normal incidence) to π/2 (grazing which ΓT M = 0. incidence). We observe: In non-magnetic media with ǫr1 < ǫr2 , ΓT M is real-valued and increases from a negative value for normal incidence to +1 as ψ i approaches grazing incidence. Note that at any particular angle of incidence, ΓT M trends toward −1 as ǫr2 /ǫr1 → ∞. In this respect, the behavior of the TM component is similar to that of the TE component. In other words: As

Brewster’s angle is also referred to as the polarizing angle. The motivation for the term “polarizing angle” is demonstrated in Figure 5.23. In this figure, a plane i wave is incident with ψ i = ψB . The wave may contain TE and TM components in any combination. Applying the principle of superposition, we may consider these components separately. The TE component of the incident wave will scatter as reflected and transmitted waves which are also TE. i However, ΓT M = 0 when ψ i = ψB , so the TM

5.10. TM REFLECTION IN NON-MAGNETIC MEDIA component of the transmitted wave will be TM, but the TM component of the reflected wave will be zero. Thus, the total (TE+TM) reflected wave will be purely TE, regardless of the TM component of the incident wave. This principle can be exploited to suppress the TM component of a wave having both TE and TM components. This method can be used to isolate the TE and TM components of a wave.

87

c C. Wang CC BY-SA 4.0

Derivation of a formula for Brewster’s angle. Figure 5.24: Derivation of a simple formula for BrewBrewster’s angle for any particular combination of ster’s angle for non-magnetic media. non-magnetic media may be determined as follows. ΓT M = 0 when the numerator of Equation 5.241 Thus, we have found equals zero, so: q r i i =0 ǫr2 − R cos ψB + R − sin2 ψB (5.243) i tan ψB = (5.253) ǫr1 where we have made the substitution R , ǫr2 /ǫr1 to improve clarity. Moving the second term to the right side of the equation and squaring both sides, we Example 5.10. Polarizing angle for an obtain: air-to-glass interface. i i R2 cos2 ψB = R − sin2 ψB

(5.244)

Now employing a trigonometric identity on the left side of the equation, we obtain:  i i R2 1 − sin2 ψB = R − sin2 ψB (5.245) i i R2 − R2 sin2 ψB = R − sin2 ψB  i 1 − R2 sin2 ψB = R − R2

and finally

(5.246) (5.247)

r

A plane wave is incident from air onto the planar boundary with a glass region. The glass exhibits relative permittivity of 2.1. The incident wave contains both TE and TM components. At what angle of incidence ψ i will the reflected wave be purely TE? Solution. Using Equation 5.253: r r ǫr2 2.1 ∼ i tan ψB = = = 1.449 ǫr1 1

R − R2 (5.254) (5.248) 1 − R2 Although this equation gets the job done, it is possible i ∼ Therefore, Brewster’s angle is ψB = 55.4◦ . This to simplify further. Note that Equation 5.248 can be i is the angle ψ at which ΓT M = 0. Therefore, interpreted as a description of the right triangle shown i when ψ i = ψB , the reflected wave contains no in Figure 5.24. In the figure, we have identified the TM component and therefore must be purely TE. length h of the vertical side and the length d of the hypotenuse as: p h , R − R2 (5.249) Additional Reading: p d , 1 − R2 (5.250) • “Brewster’s angle” on Wikipedia. The length b of the horizontal side is therefore p √ b = d2 − h2 = 1 − R (5.251) i sin ψB

=

Subsequently, we observe √ √ R − R2 h i tan ψB = = √ = R b 1−R

(5.252)

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

88

5.11 Total Internal Reflection

explicitly is: t

ψ = arcsin

[m0169]

Total internal reflection refers to a particular condition resulting in the complete reflection of a wave at the boundary between two media, with no power transmitted into the second region. One way to achieve complete reflection with zero transmission is simply to require the second material to be a perfect conductor. However, total internal reflection is a distinct phenomenon in which neither of the two media are perfect conductors. Total internal reflection has a number of practical applications; notably, it is the enabling principle of fiber optics. Consider the situation shown in Figure 5.25: A uniform plane wave is obliquely incident on the planar boundary between two semi-infinite material regions. In Section 5.8, it is found that

r

µr1 ǫr1 sin ψ i µr2 ǫr2



(5.257)

From Equation 5.257, it is apparent that when µr1 ǫr1 > µr2 ǫr2 , ψ t > ψ r ; i.e., the transmitted wave appears to bend away from the surface normal, as shown in Figure 5.25. In fact, ψ t can be as large as π/2 (corresponding to propagation parallel to the boundary) for angles of incidence which are less than π/2. What happens if the angle of incidence is further increased? When calculating ψ t using Equation 5.257, one finds that the argument of the arcsine function becomes greater than 1. Since the possible values of the sine function are between −1 and +1, the arcsine function is undefined. Clearly our analysis is inadequate in this situation.

To make sense of this, let us begin by identifying the threshold angle of incidence ψci at which the trouble ψr = ψi (5.255) begins. From the analysis in the previous paragraph, r i.e., angle of reflection equals angle of incidence. µr1 ǫr1 (5.258) sin ψci = 1 Also, from Snell’s law: µr2 ǫr2 √ √ µr1 ǫr1 sin ψ i = µr2 ǫr2 sin ψ t (5.256) therefore, r µr2 ǫr2 i ψc = arcsin (5.259) where “r” in the subscripts indicates the relative µr1 ǫr1 (unitless) quantities. The associated formula for ψ t This is known as the critical angle. When ψ i < ψci , our existing theory applies. When ψ i ≥ ψci , the situation is, at present, unclear. For non-magnetic materials, Equation 5.259 simplifies to r ǫr2 i (5.260) ψc = arcsin ǫr1

c C. Wang CC BY-SA 4.0 (modified)

Now let us examine the behavior of the reflection coefficient. For example, the reflection coefficient for TE component and non-magnetic materials is q cos ψ i − ǫr2 /ǫr1 − sin2 ψ i q ΓT E = (5.261) cos ψ i + ǫr2 /ǫr1 − sin2 ψ i

Figure 5.25: A uniform plane wave obliquely inci- From Equation 5.260, we see that dent on the planar boundary between two semi-infinite ǫr2 sin2 ψci = material regions. Here µr1 ǫr1 > µr2 ǫr2 , so ψ t > ψ i . ǫr1

(5.262)

5.11. TOTAL INTERNAL REFLECTION

89

So, when ψ i > ψci , we see that ǫr2 /ǫr1 − sin2 ψ i < 0 (ψ i > ψci ) and therefore q ǫr2 /ǫr1 − sin2 ψ i = jB (ψ i > ψci )

(5.263)

(5.264)

where B is a positive real-valued number. Now we may write Equation 5.261 as follows: ΓT E =

A − jB (ψ i > ψci ) A + jB

(5.265)

where A , cos ψ i is also a positive real-valued number. Note that the numerator and denominator in Equation 5.265 have equal magnitude. Therefore, the magnitude of |ΓT E | = 1 and ΓT E = ejζ (ψ i > ψci )

(5.266)

where ζ is a real-valued number indicating the phase of ΓT E . In other words,

c Z. S´andor CC BY-SA 3.0

Figure 5.26: Total internal reflection of a light wave incident on a planar boundary between glass and air.

When the angle of incidence ψ i exceeds the critical angle ψci , the magnitude of the reflection coefficient is 1. In this case, all power is reflected, and no power is transmitted into the second medium. This is total internal reflection. Although we have obtained this result for TE component, the identical conclusion is obtained for TM component as well. This is left as an exercise for the student. Example 5.11. Total internal reflection in glass. Figure 5.15 (Section 5.8) shows a demonstration of refraction of a beam of light incident from air onto a planar boundary with glass. Analysis of that demonstration revealed that the relative permittivity of the glass was ∼ = 2.28. Figure 5.26 shows a modification of the demonstration: In this case, a beam of light incident from glass onto a planar boundary with air. Confirm that total internal reflection is the expected result in this demonstration. Solution. Assuming the glass is non-magnetic, the critical angle is given by Equation 5.260. In the present example, ǫr1 ∼ = 2.28 (glass), ǫr2 ∼ =1 (air). Therefore, the critical angle ψci ∼ = 41.5◦ .

c Timwether CC BY-SA 3.0

Figure 5.27: Laser light in a dielectric rod exhibiting the Goos-H¨anchen effect. The angle of incidence in Figure 5.26 is seen to be about ∼ = 50◦ , which is greater than the critical angle. Therefore, total internal reflection is expected.

Note also that the phase ζ of the reflection coefficient is precisely zero unless ψ i > ψci , at which point it is both non-zero and varies with ψ i . This is known as the Goos-H¨anchen effect. This leads to the startling phenomenon shown in Figure 5.27. In this figure, the Goos-H¨anchen phase shift is apparent as a displacement between the point of incidence and the point of reflection.

90

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

The presence of an imaginary component in the reflection coefficient is odd for two reasons. First, we are not accustomed to seeing a complex-valued reflection coefficient emerge when the wave impedances of the associated media are real-valued. Second, the total reflection of the incident wave seems to contradict the boundary condition that requires the tangential components of the electric and magnetic fields to be continuous across the boundary. That is, how can these components of the fields be continuous across the boundary if no power is transmitted across the boundary? These considerations suggest that there is a field on the opposite side of the boundary, but – somehow – it must have zero power. This is exactly the case, as is explained in Section 5.12. Additional Reading: • “Goos-H¨anchen effect” on Wikipedia. • “Snell’s law” on Wikipedia. • “Total internal reflection” on Wikipedia.

5.12 Evanescent Waves [m0170]

Consider the situation shown in Figure 5.28: A uniform plane wave obliquely incident on the planar boundary between two semi-infinite material regions, and total internal reflection occurs because the angle of incidence ψ i is greater than the critical angle r µr2 ǫr2 i (5.267) ψc = arcsin µr1 ǫr1 Therefore, the reflection coefficient is complex-valued with magnitude equal to 1 and phase that depends on polarization and the constitutive parameters of the media. The total reflection of the incident wave seems to contradict the boundary conditions that require the tangential components of the electric and magnetic fields to be continuous across the boundary. How can these components of the fields be continuous across the boundary if no power is transmitted across the boundary? There must be a field on the opposite side of the boundary, but – somehow – it must have zero power. To make sense of this, let us attempt to find a solution for the transmitted field.

c C. Wang CC BY-SA 4.0

Figure 5.28: A uniform plane wave obliquely incident on the planar boundary between two semi-infinite material regions. Here µr1 ǫr1 > µr2 ǫr2 and ψ i > ψci .

5.12. EVANESCENT WAVES

91

We begin by postulating a complex-valued angle of transmission ψ tc . Although the concept of a complex-valued angle may seem counterintuitive, there is mathematical support for this concept. For example, consider the well-known trigonometric identities:  1 jθ e − e−jθ j2  1 jθ cos θ = e + e−jθ 2 sin θ =

(5.268)

Based on the evidence established so far, we presume that ψ tc behaves as follows:

ψ

tc

ψ

tc

, π/2 ,

ψ i < ψci i

ψ = ′′

ψci i

, π/2 + jψ , ψ >

Using Equation 5.268, we find:  ′′ 1  j (jψ′′ ) e − e−j (jψ ) j2  ′′ 1  −ψ′′ = e − e+ψ j2  ′′ 1  +ψ′′ e − e−ψ =j 2 = j sinh ψ ′′

sin jψ ′′ =

(5.269)

These identities allow us to compute values for sine and cosine even when θ is complex-valued. One may conclude that the sine and cosine of a complex-valued angle exist, although the results may also be complex-valued.

ψ tc , ψ t ,

In other words, the cosine of jψ ′′ is simply the hyperbolic cosine (“cosh”) of ψ ′′ . Interestingly, cosh of a real-valued argument is real-valued, so cos jψ ′′ is real-valued.

(5.270) (5.271) ψci

(5.272)

In other words, ψ tc is identical to ψ t for ψ i ≤ ψci , but when total internal reflection occurs, we presume the real part of ψ tc remains fixed (parallel to the boundary) and that an imaginary component jψ ′′ emerges to satisfy the boundary conditions. ′

For clarity, let us assign the variable ψ to represent the real part of ψ tc in Equations 5.270–5.272. Then we may refer to all three cases using a single expression as follows: ψ tc = ψ ′ + jψ ′′

sin ψ tc = sin (ψ ′ + jψ ′′ ) = sin ψ ′ cos jψ ′′ + cos ψ ′ sin jψ ′′ Using Equation 5.269, we find:  ′′ 1  j (jψ′′ ) cos jψ ′′ = e + e−j (jψ ) 2  ′′ 1  −ψ′′ = e + e+ψ 2 = cosh ψ ′′

(5.274) (5.275)

(5.277) (5.278)

(5.281) (5.282)

Using these results, we find Equation 5.275 may be written as follows: sin ψ tc = sin ψ ′ cosh ψ ′′ + j cos ψ ′ sinh ψ ′′ (5.283) Using precisely the same approach, we find: cos ψ tc = cos ψ ′ cosh ψ ′′ − j sin ψ ′ sinh ψ ′′ (5.284) Before proceeding, let’s make sure Equations 5.283 and 5.284 exhibit the expected behavior before the onset of total internal reflection. For ψ i < ψci , ψ ′ = ψ t and ψ ′′ = 0. In this case, sinh ψ ′′ = 0, cosh ψ ′′ = 1, and Equations 5.283 and 5.284 yield sin ψ tc = sin ψ t cos ψ

tc

= cos ψ

(5.285)

t

(5.286)

as expected. When total internal reflection is in effect, ψ i > ψci , so ψ ′ = π/2. In this case, Equations 5.283 and 5.284 yield sin ψ tc = cosh ψ ′′

(5.276)

(5.280)

In other words, the sine of jψ ′′ is j times hyperbolic sine (“sinh”) of ψ ′′ . Now note that sinh of a real-valued argument is real-valued, so sin jψ ′′ is imaginary-valued.

(5.273)

Now we use a well-known trigonometric identity as follows:

(5.279)

cos ψ

tc

= −j sinh ψ

(5.287) ′′

(5.288)

Let us now consider what this means for the field in Region 2. According to the formalism adopted in previous sections, the propagation of wave

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

92

components in this region is described by the factor t e−jk ·r where ˆt k t = β2 k ˆ sin ψ tc + z ˆ cos ψ tc = β2 x and



ˆx + y ˆy + z ˆz r=x so   kt · r = β2 sin ψ tc x + β2 cos ψ tc z

evanescent wave. It suffices to say that the magnitude of the evanescent field becomes negligible beyond a few wavelengths of the boundary.

(5.289) Finally, we return to the strangest characteristic of this field: It acts like a wave, but conveys no power. This field exists solely to enforce the electromagnetic (5.290) boundary conditions at the boundary, and does not exist independently of the incident and reflected field. The following thought experiment may provide some additional insight.

= (β2 cosh ψ ′′ ) x + (−jβ2 sinh ψ ′′ ) z (5.291) In this experiment, a laser illuminates a planar boundary between two material regions, with Therefore, the wave in Region 2 propagates according conditions such that total internal reflection occurs. to Thus, all incident power is reflected, and an evanescent wave exists on the opposite side of the ′′ ′′ t e−jk ·r = e−j (β2 cosh ψ )x e−(β2 sinh ψ )z (5.292) boundary. Next, the laser is turned off. All the light incident on the boundary reflects from the boundary Note that the constants β2 cosh ψ ′′ and β2 sinh ψ ′′ are and continues to propagate to infinity, even after light both real-valued and positive. Therefore, is no longer incident on the boundary. In contrast, the Equation 5.292 describes a wave which propagates in evanescent wave vanishes at the moment laser light the +ˆ x direction, but which is not uniform. ceases to illuminate the boundary. In other words, the Specifically, the magnitude of the transmitted field evanescent field does not continue to propagate along decreases exponentially with increasing z; i.e., the boundary to infinity. The reason for this is simply maximum at the boundary, asymptotically that there is no power – hence, no energy – in the approaching zero with increasing distance from the evanescent wave. boundary. This wave is unlike the incident or reflected waves (both uniform plane waves), and is Additional Reading: unlike the transmitted wave in the ψ i < ψci case (also a uniform plane wave). The transmitted wave that we • “Evanescent field” on Wikipedia. have derived in the ψ i > ψci case gives the impression of being somehow attached to the boundary, and so • “Hyperbolic function” on Wikipedia. may be described as a surface wave. However, in this case we have a particular kind of surface wave, • “Total internal reflection” on Wikipedia. known as an evanescent wave. Summarizing: When total internal reflection occurs, the transmitted field is an evanescent wave; i.e., a surface wave which conveys no power and whose magnitude decays exponentially with increasing distance into Region 2. At this point, we could enforce the “phase matching” condition at the boundary, which would lead us to a new version of Snell’s law that would allow us to solve for ψ ′′ in terms of ψ i and the constitutive properties of the media comprising Regions 1 and 2. We could subsequently determine values for the phase propagation and attenuation constants for the

[m0185]

5.12. EVANESCENT WAVES

93

Image Credits c Sevenchw (C. Wang), Fig. 5.1: https://commons.wikimedia.org/wiki/File:Upw incident on planar boundary.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), https://commons.wikimedia.org/wiki/File:Upw incident on a slab.svg, Fig. 5.2: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 5.3: https://commons.wikimedia.org/wiki/File:Double-boundary problem equivalent.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), https://commons.wikimedia.org/wiki/File:Plane wave in basic coord.svg, Fig. 5.4: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), https://commons.wikimedia.org/wiki/File:Plane wave in rotated coord.svg, Fig. 5.5: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 5.6: https://commons.wikimedia.org/wiki/File:Plane wave in another rotation coord.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 5.7: https://commons.wikimedia.org/wiki/File:Plane wave in ray-fixed coord.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 5.8: https://commons.wikimedia.org/wiki/File:Coord system for TE-TM decomposition.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 5.9: https://commons.wikimedia.org/wiki/File:TE-polarized upw incident on planar boundary.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 5.10: https://commons.wikimedia.org/wiki/File:Incident reflected and transmitted magnetic field components.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 5.11: https://commons.wikimedia.org/wiki/File:TE-polarized upw incident from air to glass.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 5.12: https://commons.wikimedia.org/wiki/File:TM-polarized upw incident on planar boundary.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 5.13: https://commons.wikimedia.org/wiki/File:TM-polarized upw incident from air to glass.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 5.14: https://commons.wikimedia.org/wiki/File:Upw incident obliquely on planar boundary.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/).

94

CHAPTER 5. WAVE REFLECTION AND TRANSMISSION

c Z. S´andor, https://commons.wikimedia.org/wiki/File:F%C3%A9nyt%C3%B6r%C3%A9s.jpg, Fig. 5.15: CC BY-SA 3.0 (https://creativecommons.org/licenses/by-sa/3.0/). c G. Saini, https://kids.kiddle.co/Image:Refractionn.jpg, Fig. 5.16: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c D-Kuru, https://commons.wikimedia.org/wiki/File:Prism-side-fs PNr%C2%B00117.jpg, Fig. 5.17: CC BY-SA 3.0 (https://creativecommons.org/licenses/by-sa/3.0/). c Suidroot, https://commons.wikimedia.org/wiki/File:Prism-rainbow.svg, Fig. 5.18: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 5.19: https://commons.wikimedia.org/wiki/File:TE-polarized upw incident on planar boundary.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 5.21: https://commons.wikimedia.org/wiki/File:TM-polarized upw incident on planar boundary.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 5.23: https://commons.wikimedia.org/wiki/File:Reflection of plane wave incidence angle equals polarizing angle.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 5.24: https://commons.wikimedia.org/wiki/File:Brewster%E2%80%99s angle for non-magnetic media.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 5.25: https://commons.wikimedia.org/wiki/File:Upw incident on planar boundary angle reflection equals incidence.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), modified. c Z. S´andor, Fig. 5.26: https://en.wikipedia.org/wiki/File:Teljes f%C3%A9nyvisszaver%C5%91d%C3%A9s.jpg, CC BY-SA 3.0 (https://creativecommons.org/licenses/by-sa/3.0/). c Timwether, https://en.wikipedia.org/wiki/File:Laser in fibre.jpg, Fig. 5.27: CC BY-SA 3.0 (https://creativecommons.org/licenses/by-sa/3.0/). c Sevenchw (C. Wang), Fig. 5.28: https://commons.wikimedia.org/wiki/File:Upw incident on planar boundary total internal reflection.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/).

Chapter 6

Waveguides 6.1 Phase and Group Velocity

Central to the concept of phase velocity is uniformity over space and time. Equations 6.2–6.4 presume a wave having the form of Equation 6.1, which exhibits [m0176] precisely the same behavior over all possible time t from −∞ to +∞ and over all possible z from −∞ to Phase velocity is the speed at which a point of +∞. This uniformity over all space and time constant phase travels as the wave propagates.1 For a precludes the use of such a wave to send information. sinusoidally-varying wave, this speed is easy to To send information, the source of the wave needs to quantify. To see this, consider the wave: vary at least one parameter as a function of time; for A cos (ωt − βz + ψ) (6.1) example A (resulting in amplitude modulation), ω (resulting in frequency modulation), or ψ (resulting in phase modulation). In other words, information can where ω = 2πf is angular frequency, z is position, and β is the phase propagation constant. At any given be transmitted only by making the wave non-uniform time, the distance between points of constant phase is in some respect. Furthermore, some materials and one wavelength λ. Therefore, the phase velocity vp is structures can cause changes in ψ or other combinations of parameters which vary with position vp = λf (6.2) or time. Examples include dispersion and propagation within waveguides. Regardless of the cause, varying Since β = 2π/λ, this may also be written as follows: the parameters ω or ψ as a function of time means that the instantaneous distance between points of ω vp = (6.3) constant phase may be very different from λ. Thus, β the instantaneous frequency of variation as a function √ of time and position may be very different from f . In Noting that β = ω µǫ for simple matter, we may this case Equations 6.2–6.4 may not necessarily also express vp in terms of the constitutive parameters provide a meaningful value for the speed of µ and ǫ as follows: propagation. 1 vp = √ (6.4) Some other concept is required to describe the speed µǫ of propagation of such waves. That concept is group Since vp in this case depends only on the constitutive velocity, vg , defined as follows: properties µ and ǫ, it is reasonable to view phase velocity also as a property of matter. 1 Formally, “velocity” is a vector which indicates both the direction and rate of motion. It is common practice to use the terms “phase velocity” and “group velocity” even though we are actually referring merely to rate of motion. The direction is, of course, in the direction of propagation.

Group velocity, vg , is the ratio of the apparent change in frequency ω to the associated change in the phase propagation constant β; i.e., ∆ω/∆β.

c 2020 S.W. Ellingson CC BY SA 4.0. https://doi.org/10.21061/electromagnetics-vol-2 Electromagnetics Vol. 2.

CHAPTER 6. WAVEGUIDES

96 Letting ∆β become vanishingly small, we obtain vg ,

∂ω ∂β

Solution. First, note

Note the similarity to the definition of phase velocity in Equation 6.3. Group velocity can be interpreted as the speed at which a disturbance in the wave propagates. Information may be conveyed as meaningful disturbances relative to a steady-state condition, so group velocity is also the speed of information in a wave. Note Equation 6.5 yields the expected result for waves in the form of Equation 6.1: −1  −1 ∂ √ ∂β = ω µǫ vg = ∂ω ∂ω 1 = √ = vp µǫ

√ √ √ β = ω µ0 ǫ = ω µ0 ǫ0 ǫr √ K · ω2 √ = µ0 ǫ0 ω0

(6.5)

The phase velocity is: vp =

ω0 ω =√ √ β K · ω µ0 ǫ0

(6.6)

In other words, the group velocity of a wave in the form of Equation 6.1 is equal to its phase velocity. To observe a difference between vp and vg , β must somehow vary as a function of something other than just ω and the constitutive parameters. Again, modulation (introduced by the source of the wave) and dispersion (frequency-dependent constitutive parameters) are examples in which vg is not necessarily equal to vp . Here’s an example involving dispersion: Example 6.1. Phase and group velocity for a material exhibiting square-law dispersion. A broad class of non-magnetic dispersive media exhibit relative permittivity ǫr that varies as the square of frequency over a narrow range of frequencies centered at ω0 . For these media we presume  2 ω ǫr = K (6.7) ω0 where K is a real-valued positive constant. What is the phase and group velocity for a sinusoidally-varying wave in this material?

(6.9)

Whereas the group velocity is:  −1 ∂β ∂ω = ∂β ∂ω !−1 √ ∂ K · ω2 √ = µ0 ǫ0 ∂ω ω0 !−1 √ K · ω√ µ0 ǫ0 = 2 ω0

vg =



(6.8)

(6.10) (6.11)

(6.12)

Now simplifying using Equation 6.8: 

β vg = 2 ω 1ω = 2β 1 = vp 2

−1

(6.13) (6.14) (6.15)

Thus, we see that in this case the group velocity is always half the phase velocity. Another commonly-encountered example for which vg is not necessarily equal to vp is the propagation of guided waves; e.g., waves within a waveguide. In fact, such waves may exhibit phase velocity greater than the speed of light in a vacuum, c. However, the group velocity remains less than c, which means the speed at which information may propagate in a waveguide is less than c. No physical laws are violated, since the universal “speed limit” c applies to information, and not simply points of constant phase. (See “Additional Reading” at the end of this section for more on this concept.) Additional Reading: • “Group velocity” on Wikipedia.

6.2. PARALLEL PLATE WAVEGUIDE: INTRODUCTION • “Phase velocity” on Wikipedia. • “Speed of light” on Wikipedia.

97

6.2 Parallel Plate Waveguide: Introduction [m0173]

A parallel plate waveguide is a device for guiding the propagation of waves between two perfectly-conducting plates. Our primary interest in this structure is as a rudimentary model applicable to a broad range of engineering problems. Examples of such problems include analysis of the fields within microstrip line and propagation of radio waves in the ionosphere. Figure 6.1 shows the geometry of interest. Here the plates are located at z = 0 and z = a. The plates are assumed to be infinite in extent, and therefore there is no need to consider fields in the regions z < 0 or z > a. For this analysis, the region between the plates is assumed to consist of an ideal (lossless) material exhibiting real-valued permeability µ and real-valued permittivity ǫ. Let us limit our attention to a region within the waveguide which is free of sources. Expressed in phasor form, the electric field intensity is governed by the wave equation

where

e + β2E e =0 ∇2 E √ β = ω µǫ

(6.16) (6.17)

Equation 6.16 is a partial differential equation. This equation, combined with boundary conditions

c C. Wang CC BY-SA 4.0 (modified)

Figure 6.1: Geometry for analysis of fields in a parallel plate waveguide.

CHAPTER 6. WAVEGUIDES

98 imposed by the perfectly-conducting plates, is sufficient to determine a unique solution. This is most easily done in Cartesian coordinates, as we shall now e in Cartesian demonstrate. First we express E coordinates: e =x ex + y ey + z ez ˆE ˆE ˆE E

e (Equation 6.18) is the solution to simultaneous E Equations 6.26–6.28 subject to the boundary conditions that apply at the PEC surfaces at z = 0 and e is z = d; namely, that the tangent component of E zero at these surfaces.

(6.18) At this point, the problem has been reduced to a routine exercise in the solution of partial differential This facilitates the decomposition of Equation 6.16 equations. However, a somewhat more useful ˆ, y ˆ, z ˆ into separate equations governing the x approach is to first decompose the total electric field e components of E: into transverse electric (TE) and transverse magnetic ex + β 2 E ex = 0 ∇2 E (6.19) (TM) components,2 and determine the solutions to these components separately. The TE component of ey + β 2 E ey = 0 ∇2 E (6.20) the electric field is parallel to the plates, and therefore ez + β 2 E ez = 0 ∇2 E (6.21) transverse (perpendicular) to the plane shown in Figure 6.1. Thus, the TE component has ex = E ez = 0, and only E ey may be non-zero. The 2 E Next we observe that the operator ∇ may be e is TM component of the magnetic field intensity (H) expressed in Cartesian coordinates as follows: parallel to the plates, and therefore transverse to the ∂2 ∂2 ∂2 2 plane shown in Figure 6.1. Thus, the TM component + 2+ 2 (6.22) ∇ = ex = H e z = 0, and only H e y may be non-zero. As ∂x2 ∂y ∂z has H always, the total field is the sum of the TE and TM so the equations governing the Cartesian components components. e may be written as follows: of E ∂2 e ∂2 e ∂2 e ex Ex + 2 E Ex = −β 2 E x+ 2 ∂x ∂y ∂z 2 ∂2 e ∂2 e ∂2 e ey Ey + 2 E Ey = −β 2 E y + 2 ∂x ∂y ∂z 2 ∂2 e ∂2 e ∂2 e ez Ez + 2 E Ez = −β 2 E z + 2 ∂x ∂y ∂z 2

The TE and TM solutions are presented in Sections 6.3 (Electric component of the TE solution), 6.4 (Magnetic component of the TE solution), and 6.5 (6.24) (Electric component of the TM solution). The magnetic component of the TM solution can be (6.25) determined via a straightforward variation of the preceding three cases, and so is not presented here. (6.23)

Let us restrict our attention to scenarios that can be completely described in two dimensions; namely x and z, and for which there is no variation in y. This is not necessarily required, however, it is representative of a broad class of relevant problems, and allows Equations 6.23–6.25 to be considerably simplified. If there is no variation in y, then partial derivatives with respect to y are zero, yielding: ∂2 e Ex + ∂x2 ∂2 e Ey + ∂x2 ∂2 e Ez + ∂x2

∂2 e ex Ex = −β 2 E ∂z 2 ∂2 e ey Ey = −β 2 E ∂z 2 ∂2 e ez Ez = −β 2 E ∂z 2

Presentation of the solution for the fields in a parallel plate waveguide under the conditions described in the section continues in Section 6.3.

(6.26) (6.27) (6.28)

The solution to the parallel plate waveguide problem may now be summarized as follows: The electric field

2 For

a refresher on TE-TM decomposition, see Section 5.5.

6.3. PARALLEL PLATE WAVEGUIDE: TE CASE, ELECTRIC FIELD

6.3 Parallel Plate Waveguide: TE Case, Electric Field [m0174]

In Section 6.2, the parallel plate waveguide was introduced. At the end of that section, we described the decomposition of the problem into its TE and TM components. In this section, we find the electric field component of the TE field in the waveguide. Figure 6.2 shows the problem addressed in this section. (Additional details and assumptions are addressed in Section 6.2.) ex = E ez = 0 for the TE component of the Since E electric field, Equations 6.26 and 6.28 are irrelevant, leaving only: ∂2 e ∂2 e ey E + Ey = −β 2 E (6.29) y ∂x2 ∂z 2 The general solution to this partial differential equation is:   ey = e−jkz z Ae−jkx x + Be+jkx x E   + e+jkz z Ce−jkx x + De+jkx x (6.30)

99

Note that Equation 6.30 consists of two terms. The first term includes the factor e−jkz z , indicating a wave propagating in the +ˆ z direction, and the second term includes the factor e+jkz z , indicating a wave propagating in the −ˆ z direction. If we impose the restriction that sources exist only on the left (z < 0) side of Figure 6.2, and that there be no structure capable of wave scattering (in particular, reflection) on the right (z > 0) side of Figure 6.2, then there can be no wave components propagating in the −ˆ z direction. In this case, C = D = 0 and Equation 6.30 simplifies to:   ey = e−jkz z Ae−jkx x + Be+jkx x E (6.31) Before proceeding, let’s make sure that Equation 6.31 is actually a solution to Equation 6.29. As we shall see in a moment, performing this check will reveal some additional useful information. First, note:

So:

ey   ∂E = e−jkz z −Ae−jkx x + Be+jkx x (jkx ) ∂x (6.32)

ey    ∂2E = e−jkz z Ae−jkx x + Be+jkx x −kx2 ∂x2 (6.33) Comparing this to Equation 6.31, we observe the remarkable fact that

where A, B, C, and D are complex-valued constants and kx and kz are real-valued constants. We have ey ∂2E assigned variable names to these constants with ey = −kx2 E (6.34) advance knowledge of their physical interpretation; ∂x2 however, at this moment they remain simply unknown Similarly, note: constants whose values must be determined by ey   ∂E enforcement of boundary conditions. = e−jkz z Ae−jkx x + Be+jkx x (−jkz ) ∂z (6.35) So: ey    ∂2E = e−jkz z Ae−jkx x + Be+jkx x −kz2 2 ∂z ey = −kz2 E (6.36)

Now summing these results:

ey ey  ∂2E ∂2E ey + = − kx2 + kz2 E (6.37) ∂x2 ∂z 2 Comparing Equation 6.37 to Equation 6.29, we c C. Wang CC BY-SA 4.0

conclude that Equation 6.31 is indeed a solution to Figure 6.2: TE component of the electric field in a Equation 6.29, but only if: parallel plate waveguide. β 2 = kx2 + kz2 (6.38)

CHAPTER 6. WAVEGUIDES

100 This confirms that kx and kz are in fact the components of the propagation vector ˆ=x ˆ kx + y ˆ ky + z ˆ kz k , βk

(6.39)

ˆ is the unit vector pointing in the direction of where k propagation, and ky = 0 in this particular problem.

from the solution for the corresponding positive value of m in sign only, which can be absorbed in the arbitrary constant Ey0 . At this point we have uncovered a family of solutions given by Equation 6.43 and Equation 6.46 with m = 1, 2, .... Each solution associated with a particular value of m is referred to as a mode, which (via Equation 6.46) has a particular value of kx . The value of kz for mode m is obtained using Equation 6.38 as follows: p kz = β 2 − kx2 r  mπ 2 (6.47) = β2 − a

The solution has now been reduced to finding the constants A, B, and either kx or kz . This is accomplished by enforcing the relevant boundary e that is conditions. In general, the component of E tangent to a perfectly-conducting surface is zero. ey = 0 at Applied to the present problem, this means E e x = 0 and Ey = 0 at x = a. Referring to Equation 6.31, the boundary condition at x = 0 means Since kz is specified to be real-valued, we require: e−jkz z [A · 1 + B · 1] = 0 (6.40)  mπ 2 The factor e−jkz z cannot be zero; therefore, β2 − >0 (6.48) a A + B = 0. Since B = −A, we may rewrite Equation 6.31 as follows:   ey = e−jkz z B e+jkx x − e−jkx x E

This constrains β; specifically:

(6.41)

This expression is simplified using a trigonometric identity:  1  +jkx x e − e−jkx x = sin kx x 2j

Now applying the boundary condition at x = a: Ey0 e−jkz z sin kx a = 0

mπ a

(6.49)

√ Recall that β = ω µǫ and ω = 2πf where f is frequency. Solving for f , we find:

(6.42) f>

Let us now make the definition Ey0 , j2B. Then: ey = Ey0 e−jkz z sin kx x E

β>

m √ 2a µǫ

(6.50)

Therefore, each mode exists only above a certain (6.43) frequency, which is different for each mode. This cutoff frequency fc for mode m is given by

(6.44)

fc(m) ,

m √ 2a µǫ

(6.51)

The factor e−jkz z cannot be zero, and Ey0 = 0 yields only trivial solutions; therefore: sin kx a = 0 This in turn requires that kx a = mπ

At frequencies below the cutoff frequency for mode m, modes 1 through m − 1 exhibit imaginary-valued (6.45) kz . The propagation constant must have a real-valued component in order to propagate; therefore, these modes do not propagate and may be ignored.

(6.46) Let us now summarize the solution. For the scenario depicted in Figure 6.2, the electric field component of where m is an integer. Note that m = 0 is not of the TE solution is given by: interest since this yields kx = 0, which according to ey = 0. Also Equation 6.43 yields the trivial solution E ∞ X ey = y e (m) each integer value of m that is less than zero is ˆE ˆ y E (6.52) y excluded because the associated solution is different m=1

6.3. PARALLEL PLATE WAVEGUIDE: TE CASE, ELECTRIC FIELD

101

where ey(m) E

,

(

0, (m) (m) (m) Ey0 e−jkz z sin kx x,

(m)

f < fc (m) f ≥ fc

(6.53)

where m enumerates modes (m = 1, 2, ...), r h i kz(m) ,

(m)

β 2 − kx

2

(6.54) c C. Wang CC BY-SA 4.0

and kx(m) , mπ/a

(6.55)

(m)

Finally, Ey0 is a complex-valued constant that depends on sources or boundary conditions to the left of the region of interest. For the scenario depicted in Figure 6.2, the electric field component of the TE solution is given by Equation 6.52 with modal components determined as indicated by Equations 6.51–6.55. This solution presumes all sources lie to the left of the region of interest, and no scattering occurs to the right of the region of interest. To better understand this result, let us examine the lowest-order mode, m = 1. For this mode √ (1) fc = 1/2a µǫ, so this mode can exist if √ (1) f > 1/2a µǫ. Also kx = π/a, so r  π 2 (1) (6.56) kz = β 2 − a Subsequently,

ey(1) = E (1) e−jkz(1) z sin πx E y0 a

(6.57)

Note that this mode has the form of a plane wave. The plane wave propagates in the +ˆ z direction with phase (1) propagation constant kz . Also, we observe that the apparent plane wave is non-uniform, exhibiting magnitude proportional to sin πx/a within the waveguide. This is shown in Figure 6.3 (left image). In particular, we observe that the magnitude of the wave is zero at the perfectly-conducting surfaces – as is necessary to satisfy the boundary conditions – and is maximum in the center of the waveguide. Now let us examine the m = 2 mode. For this mode, √ (2) fc = 1/a µǫ, so this mode can exist if

Figure 6.3: Magnitude of the two lowest-order TE modes in a parallel plate waveguide. Left: m = 1, Right: m = 2. √ (1) f > 1/a µǫ. This frequency is higher than fc , so the m = 1 mode can exist at any frequency at which (2) the m = 2 mode exists. Also kx = 2π/a, so s  2 2π (2) 2 (6.58) kz = β − a Subsequently, e (2) = E (2) e−jkz(2) z sin 2πx E (6.59) y y0 a In this case, the apparent plane wave propagates in the (2) +ˆ z direction with phase propagation constant kz , (1) which is less than kz . For m = 2, we find magnitude is proportional to sin 2πx/a within the waveguide (Figure 6.3, right image). As in the m = 1 case, we observe that the magnitude of the wave is zero at the PEC surfaces; however, for m = 2, there are two maxima with respect to x, and the magnitude in the center of the waveguide is zero. This pattern continues for higher-order modes. In particular, each successive mode exhibits higher cutoff frequency, smaller propagation constant, and increasing integer number of sinusoidal half-periods in magnitude. Example 6.2. Single-mode TE propagation in a parallel plate waveguide. Consider an air-filled parallel plate waveguide consisting of plates separated by 1 cm. Determine the frequency range for which one

CHAPTER 6. WAVEGUIDES

102 (and only one) propagating TE mode is assured. Solution. Single-mode TE propagation is assured by limiting frequency f to greater than the cutoff frequency for m = 1, but lower than the cutoff frequency for m = 2. (Any frequency higher than the cutoff frequency for m = 2 allows at least 2 modes to exist.) Calculating the applicable cutoff frequencies, we find: 1 √ 2a µ0 ǫ0 2 = √ 2a µ0 ǫ0

fc(1) =

∼ = 15.0 GHz

(6.60)

fc(2)

∼ = 30.0 GHz

(6.61)

Therefore, 15.0 GHz ≤ f ≤ 30.0 GHz. Finally, let us consider the phase velocity vp within the waveguide. For lowest-order mode m = 1, this is vp =

ω (1) kz

ω

=q

ω 2 µǫ −

 π 2 a

6.4 Parallel Plate Waveguide: TE Case, Magnetic Field [m0175]

In Section 6.2, the parallel plate waveguide was introduced. In Section 6.3, we determined the TE component of the electric field. In this section, we determine the TE component of the magnetic field. The reader should be familiar with Section 6.3 before attempting this section. In Section 6.3, the TE component of the electric field was determined to be: ey = y ˆE ˆ y

ey(m) E

(6.63)

e = −jωµH e ∇×E

(6.64)

m=1

The TE component of the magnetic field may be obtained from the Maxwell-Faraday equation:

(6.62)

Recall that the speed of an electromagnetic wave in √ unbounded space (i.e., not in a waveguide) is 1/ µǫ. For example, the speed of light in free space is √ 1/ µ0 ǫ0 = c. However, the phase velocity indicated √ by Equation 6.62 is greater than 1/ µǫ; e.g., faster than light would travel in the same material (presuming it were transparent). At first glance, this may seem to be impossible. However, recall that information travels at the group velocity vg , and not necessarily the phase velocity. (See Section 6.1 for a refresher.) Although we shall not demonstrate this here, the group velocity in the parallel plate √ waveguide is always less than 1/ µǫ, so no physical laws are broken, and signals travel somewhat slower than the speed of light, as they do in any other structure used to convey signals.

∞ X

Thus:

e e = j ∇×E H ωµ   j ey ˆE = ∇× y ωµ

(6.65)

The relevant form of the curl operator is Equation B.16 (Appendix B.2). Although the complete expression consists of 6 terms, all but 2 of ˆ and z ˆ components these terms are zero because the x e are zero. The two remaining terms are of E ey /∂z and +ˆ ey /∂x. Thus: −ˆ x∂ E z∂ E ! ey ey j ∂ E ∂ E e = ˆ H −ˆ x +z (6.66) ωµ ∂z ∂x

ey is the sum of modes, as indicated in Recall that E Equation 6.63. Since differentiation (i.e., ∂/∂z and Also remarkable is that the speed of propagation is ∂/∂x) is a linear operator, we may evaluate different for each mode. In fact, we find that the phase Equation 6.66 for modes one at a time, and then sum velocity increases and the group velocity decreases as the results. Using this approach, we find: m increases. This phenomenon is known as dispersion, and sometimes specifically as mode dispersion or modal dispersion.

(m)

ey ∂E ∂z

=

∂ (m) −jkz(m) z Ey0 e sin kx(m) x ∂z   (m)

(m)

= Ey0 e−jkz

z

sin kx(m) x

−jkz(m)



(6.67)

6.4. PARALLEL PLATE WAVEGUIDE: TE CASE, MAGNETIC FIELD and (m)

ey ∂E ∂x

(m) ∂ (m) Ey0 e−jkz z sin kx(m) x ∂x    (m) (m) = Ey0 e−jkz z cos kx(m) x +kx(m)

=

(6.68)

We may now assemble a solution for the magnetic field as follows: ex + z ez = x ˆH ˆH ˆ x

∞ X

e x(m) H

m=1 ∞ X

ˆ +z

m=1

where

e (m) H z

(6.69)

(m)

e (m) = − kz E (m) e−jkz(m) z sin k (m) x H x x ωµ y0

(6.70)

103

Recall that the magnetic field is identically zero inside a PEC material. Also recall that boundary conditions require that discontinuity in the component of H tangent to a surface must be supported by a surface current. We conclude that Current flows on the PEC surfaces of the waveguide. If this seems surprising, note that essentially the same thing happens in a coaxial transmission line. That is, signals in a coaxial transmission line can be described equally well in terms of either potentials and currents on the inner and outer conductors, or the electromagnetic fields between the conductors. The parallel plate waveguide is only slightly more complicated because the field in a properly-designed coaxial cable is a single transverse electromagnetic (TEM) mode, whereas the fields in a parallel plate waveguide are combinations of TE and TM modes.

(m)

e (m) = +j kx E (m) e−jkz(m) z cos k (m) x H z x ωµ y0

(6.71)

and modes may only exist at frequencies greater than the associated cutoff frequencies. Summarizing: The magnetic field component of the TE solution is given by Equation 6.69 with modal components as indicated by Equations 6.70 and 6.71. Caveats pertaining to the cutoff frequencies and the locations of sources and structures continue to apply. This result is quite complex, yet some additional insight is possible. At the perfectly-conducting (PEC) surface at x = 0, we see e (m) (x = 0) = 0 H x

(m)

e z(m) (x = 0) = +j kx E (m) e−jkz(m) z H ωµ y0

(6.72)

Similarly, on the PEC surface at x = a, we see e (m) (x = a) = 0 H x

(m)

e z(m) (x = a) = −j kx E (m) e−jkz(m) z H ωµ y0

(6.73)

Thus, we see the magnetic field vector at the PEC surfaces is non-zero and parallel to the PEC surfaces.

Interestingly, we also find that the magnetic field vector points in different directions depending on position relative to the conducting surfaces. We just determined that the magnetic field is parallel to the conducting surfaces at those surfaces. However, the magnetic field is perpendicular to those surfaces at m locations between x = 0 and x = a. These locations correspond to maxima in the electric field.

CHAPTER 6. WAVEGUIDES

104

6.5 Parallel Plate Waveguide: TM Case, Electric Field [m0177]

In Section 6.2, the parallel plate waveguide shown in Figure 6.4 was introduced. At the end of that section, we decomposed the problem into its TE and TM components. In this section, we find the TM component of the fields in the waveguide. “Transverse magnetic” means the magnetic field vector is perpendicular to the plane of interest, and is therefore parallel to the conducting surfaces. Thus, e =y e y , with no component in the x ˆH ˆ or z ˆ H directions. Following precisely the same reasoning employed in Section 6.2, we find the governing equation for the magnetic component of TM field is:

Note that Equation 6.75 consists of two terms. The first term includes the factor e−jkz z , indicating a wave propagating in the +ˆ z direction, and the second term includes the factor e+jkz z , indicating a wave propagating in the −ˆ z direction. If we impose the restriction that sources exist only on the left (z < 0) side of Figure 6.4, and that there be no structure capable of wave scattering (in particular, reflection) on the right (z > 0) side of Figure 6.4, then there can be no wave components propagating in the −ˆ z direction. In this case, C = D = 0 and Equation 6.75 simplifies to:   e y = e−jkz z Ae−jkx x + Be+jkx x H (6.76) Before proceeding, let’s make sure that Equation 6.76 is actually a solution to Equation 6.74. As in the TE case, this check yields a constraint (in fact, the same constraint) on the as-yet undetermined parameters kx and kz . First, note:

∂2 e ∂2 e 2 e Hy + 2 H (6.74) y = −β Hy 2 ∂x ∂z ey   ∂H The general solution to this partial differential = e−jkz z −Ae−jkx x + Be+jkx x (jkx ) ∂x equation is: (6.77)   e y = e−jkz z Ae−jkx x + Be+jkx x So: H   + e+jkz z Ce−jkx x + De+jkx x (6.75) ey    ∂2H = e−jkz z Ae−jkx x + Be+jkx x −kx2 2 ∂x where A, B, C, and D are complex-valued constants; ey = −kx2 H (6.78) and kx and kz are real-valued constants. We have assigned variable names to these constants with Next, note: advance knowledge of their physical interpretation; however, at this moment they remain simply unknown ey   ∂H constants whose values must be determined by = e−jkz z Ae−jkx x + Be+jkx x (−jkz ) ∂z enforcement of boundary conditions. (6.79) So: ey    ∂2H = e−jkz z Ae−jkx x + Be+jkx x −kz2 2 ∂z ey = −kz2 H (6.80)

Now summing these results:

ey ey  ∂2H ∂2H ey + = − kx2 + kz2 H 2 2 ∂x ∂z

c C. Wang CC BY-SA 4.0

Figure 6.4: TM component of the electric field in a parallel plate waveguide.

(6.81)

Comparing Equation 6.81 to Equation 6.74, we conclude that Equation 6.76 is a solution to Equation 6.74 under the constraint that: β 2 = kx2 + kz2

(6.82)

6.5. PARALLEL PLATE WAVEGUIDE: TM CASE, ELECTRIC FIELD This is precisely the same constraint identified in the TE case, and confirms that kx and ky are in fact the components of the propagation vector ˆ=x ˆ kx + y ˆ ky + z ˆ kz k , βk

Our objective in this section is to determine the electric field component of the TM field. The electric field may be obtained from the magnetic field using Ampere’s law:

Thus:

e = jωǫE e ∇×H e = 1 ∇×H e E jωǫ   1 ey ˆH ∇× y = jωǫ

(6.84)

(6.85)

  ez = kx e−jkz z −Ae−jkx x + Be+jkx x E ωǫ

(6.87)

(6.88)

The solution has now been reduced to the problem of finding the constants A, B, and either kx or kz . This is accomplished by enforcing the relevant boundary conditions. In general, the component of the electric field which is tangent to a perfectly-conducting surface is zero. Applied to the present (TM) case, this ez (x = 0) = 0 and E ez (x = a) = 0. means E

(6.89)

The factor e−jkz z always has unit magnitude, and so cannot be zero. We could require kx to be zero, but this is unnecessarily restrictive. Instead, we require A = B and we may rewrite Equation 6.88 as follows:   ez = Bkx e−jkz z e+jkx x − e−jkx x E ωǫ

(6.90)

This expression is simplified using a trigonometric identity:

Thus:

We may further develop this expression using ˆ component of Equations 6.77 and 6.79. We find the x e is: E e is: ˆ component of E and the z

kx −jkz z e [−A (1) + B (1)] = 0 ωǫ

sin kx a =

The relevant form of the curl operator is Equation B.16 (Appendix B.2). Although the complete expression consists of 6 terms, all but 2 of ˆ and z ˆ components these terms are zero because the x e are zero. The two remaining terms are of H e y /∂z and +ˆ e y /∂x. Thus: −ˆ x∂ H z∂ H ! ey ey 1 ∂ H ∂ H e= ˆ E −ˆ x +z (6.86) jωǫ ∂z ∂x

  ex = kz e−jkz z Ae−jkx x + Be+jkx x E ωǫ

Referring to Equation 6.88, the boundary condition at x = 0 means

(6.83)

ˆ is the unit vector pointing in the direction of where k propagation, and ky = 0 in this particular problem.

105

 1  +jkx a e − e−jkx a 2j

ez = j2Bkx e−jkz z sin kx x E ωǫ ex , beginning from Now following up with E Equation 6.87:

(6.91)

(6.92)

  ex = kz e−jkz z Ae−jkx x + Be+jkx x E ωǫ  Bkz −jkz z  −jkx x = e e + e+jkx x ωǫ 2Bkz −jkz z = e cos kx x (6.93) ωǫ

For convenience we define the following complex-valued constant: Ex0 ,

2Bkz ωǫ

(6.94)

This yields the following simpler expression: ex = Ex0 e−jkz z cos kx x E

(6.95)

sin kx a = 0

(6.97)

Now let us apply the boundary condition at x = a to ez : E j2Bkz −jkz z e sin kx a = 0 (6.96) ωǫ Requiring B = 0 or kz = 0 yields only trivial solutions, therefore, it must be true that

This in turn requires that kx a = mπ

(6.98)

CHAPTER 6. WAVEGUIDES

106 where m is an integer. Note that this is precisely the same relationship that we identified in the TE case. There is an important difference, however. In the TE case, m = 0 was not of interest because this yields kx = 0, and the associated field turned out to be identically zero. In the present (TM) case, m = 0 also yields kx = 0, but the associated field is not ez = 0 but E ex necessarily zero. That is, for m = 0, E is not necessarily zero. Therefore, m = 0 as well as m = 1, m = 2, and so on are of interest in the TM case.

may exist for any a > 0 and any f > 0. Once again, this is a remarkable difference from the TE case, for which m = 0 is not available. Let us now summarize the solution. With respect to Figure 6.4, we find that the electric field component of the TM field is given by: e= E

∞ h X

m=0

e (m) + z e (m) ˆE ˆE x x z

i

(6.104)

where At this point, we have uncovered a family of solutions ( with m = 0, 1, 2, .... Each solution is referred to as a (m) f < fc 0, (m) mode, and is associated with a particular value of kx . e Ex , (m) (m) (m) (m) Ex0 e−jkz z cos kx x, f ≥ fc In the discussion that follows, we shall find that the consequences are identical to those identified in the (6.105) TE case, except that m = 0 is now also allowed. and Continuing: The value of kz for mode m is obtained ( (m) using Equation 6.82 as follows: 0, f < fc (m) e (m) p Ez , (m) (m) (m) (m) k j x(m) Ex0 e−jkz z sin kx x, f ≥ fc kz = β 2 − kx2 kz r  mπ 2 (6.106) (6.99) where m enumerates modes (m = 0, 1, 2, ...) and = β2 − a r Since kz is specified to be real-valued, we require: h i β2 −

 mπ 2 a

This constrains β; specifically:

mπ a

kz(m) ,

>0

(6.100)

(m)

β 2 − kx

kx(m) , mπ/a (m)

2

(6.107)

(6.108)

(6.101) Finally, the coefficients Ex0 depend on sources and/or boundary conditions to the left of the region of √ interest. Recall that β = ω µǫ and ω = 2πf where f is frequency. Solving for f , we find: For the scenario depicted in Figure 6.4, the elecm tric field component of the TM solution is given f> √ (6.102) by Equation 6.104 with modal components de2a µǫ termined as indicated by Equations 6.103–6.108. Thus, each mode exists only above a certain This solution presumes all sources lie to the frequency, which is different for each mode. This left of the region of interest, with no additional cutoff frequency fc for mode m is given by sources or boundary conditions to the right of the region of interest. m fc(m) , √ (6.103) 2a µǫ The m = 0 mode, commonly referred to as the At frequencies below the cutoff frequency for mode “TM0 ” mode, is of particular importance in the m, modes 0 through m − 1 exhibit imaginary-valued analysis of microstrip transmission line, and is kz and therefore do no propagate. Also, note that the addressed in Section 6.6. cutoff frequency for m = 0 is zero, and so this mode is always able to propagate. That is, the m = 0 mode β>

6.6. PARALLEL PLATE WAVEGUIDE: THE TM0 MODE

6.6 Parallel Plate Waveguide: The TM0 Mode [m0220]

In Section 6.2, the parallel plate waveguide (also shown in Figure 6.5) was introduced. At the end of that section we decomposed the problem into its constituent TE and TM fields. In Section 6.5, we determined the electric field component of the TM field, which was found to consist of a set of discrete modes. In this section, we address the lowest-order (m = 0) mode of the TM field, which has special relevance in a number of applications including microstrip transmission lines. This mode is commonly referred to as the “TM0 ” mode. The TM electric field intensity in the waveguide is given by Equation 6.104 with modal components determined as indicated by Equations 6.103–6.108 (Section 6.5). Recall that the m = 0 mode can only exist only in the TM case, and does not exist in the TE case. We also noted that the cutoff frequency for this mode is zero, so it may exist at any frequency, and within any non-zero plate separation a. For this (0) (0) mode kx = 0, kz = β, and we find (0) e =x ˆ Ex0 e−jβz (TM0 mode) E

(6.109)

Remarkably, we find that this mode has the form of a uniform plane wave which propagates in the +ˆ z direction; i.e., squarely between the plates. The phase and group velocities of this wave are equal to each other and ω/β, precisely as expected for a uniform

107

plane wave in unbounded media; i.e., as if the plates did not exist. This observation allows us to easily determine the associated magnetic field: Using the plane wave relationship, ˆ×E e = 1k e H η   1 (0) ˆ× x ˆ Ex0 e−jβz = z η

(6.110) (6.111)

(0)

ˆ =y

Ex0 −jβz e (TM0 mode) η

(6.112)

Example 6.3. Guided waves in a printed circuit board (PCB). A very common form of PCB consists of a 1.575 mm-thick slab of low-loss dielectric having relative permittivity ≈ 4.5 sandwiched between two copper planes. Characterize the electromagnetic field in a long strip of this material. Assume a single source exists at one end of the strip, and operates at frequencies below 10 GHz. Solution. Let us assume that the copper planes exhibit conductivity that is sufficiently high that the inward-facing surfaces may be viewed as perfectly-conducting. Also, let us limit scope to the field deep within the “sandwich,” and neglect the region near the edges of the PCB. Under these conditions, the PCB is well-modeled as a parallel-plate waveguide. Thus, the electromagnetic field consists of a combination of TE and TM modes. The active (non-zero) modes depend on the source (a mode must be “stimulated” by the source in order to propagate) and modal cutoff frequencies. The cutoff frequency for mode m is fc(m) =

m √ 2a µǫ

(6.113)

In this case, a = 1.575 mm, µ ≈ µ0 , and ǫ ≈ 4.5ǫ0 . Therefore: fc(m) ≈ (44.9 GHz) m c C. Wang CC BY-SA 4.0 (modified)

Figure 6.5: TM0 mode in a parallel plate waveguide.

(6.114)

Since the cutoff frequency for m = 1 is much greater than 10 GHz, we may rest assured that

CHAPTER 6. WAVEGUIDES

108 the only mode that can propagate inside the PCB is TM0 . Therefore, the field deep inside the PCB may be interpreted as a single plane wave having the TM0 structure shown in Figure 6.5, propagating away from the source end of the PCB. The phase velocity is simply that of the apparent plane wave:

6.7 General Relationships for Unidirectional Waves [m0222]

Analysis of electromagnetic waves in enclosed spaces – and in waveguides in particular – is quite difficult. 1 vp = √ ≈ 1.41 × 108 m/s ≈ 0.47c (6.115) The task is dramatically simplified if it can be µǫ assumed that the wave propagates in a single direction; i.e., is unidirectional. This does not The scenario described in this example is essentially a necessarily entail loss of generality. For example: Within a straight waveguide, waves can travel either very rudimentary form of microstrip transmission “forward” or “backward.” The principle of line. superposition allows one to consider these two unidirectional cases separately, and then to simply sum the results. In this section, the equations that relate the various components of a unidirectional wave are derived. The equations will be derived in Cartesian coordinates, anticipating application to rectangular waveguides. However, the underlying strategy is generally applicable. We begin with Maxwell’s curl equations: e = −jωµH e ∇×E e = +jωǫE e ∇×H

(6.116) (6.117)

e Let us consider Equation 6.117 first. Solving for E, we have: e e = 1 ∇×H E jωǫ

(6.118)

This is actually three equations; that is, one each for e respectively. To ˆ, y ˆ , and z ˆ components of E, the x extract these equations, let us define the components as follows: e =x ex + y ey + z ez ˆE ˆE ˆE E e =x ex + y ey + z ez ˆH ˆH ˆH H

(6.119) (6.120)

Now applying the equation for curl in Cartesian coordinates (Equation B.16 in Appendix B.2), we

6.7. GENERAL RELATIONSHIPS FOR UNIDIRECTIONAL WAVES

109

Applying the same procedure to the curl equation for e (Equation 6.116), one obtains: H ez ey 1 ∂ H ∂ H ex = ! E − (6.121) jωǫ ∂y ∂z ez 1 ∂ E ex = ey H + jkz E (6.133) ! −jωµ ∂y ez ex 1 ∂ H ∂ H ey = ! E − (6.122) jωǫ ∂z ∂x ez 1 ∂ E ey = ex − H −jkz E (6.134) ! −jωµ ∂x ey ex 1 ∂ H ∂ H e ! Ez = − (6.123) jωǫ ∂x ∂y ey ex 1 ∂ E ∂ E ez = H − (6.135) −jωµ ∂x ∂y Without loss of generality, we may assume that the Equations 6.130–6.135 constitute a set of single direction in which the wave is traveling is in simultaneous equations that represent Maxwell’s curl e and the +ˆ z direction. If this is the case, then H, e contains a factor equations in the special case of a unidirectional subsequently each component of H, z-traveling) wave. With just a little of e−jkz z where kz is the phase propagation constant (specifically, a +ˆ bit of algebraic manipulation of these equations, it is in the direction of travel. The remaining factors are ˆ and y ˆ possible to obtain expressions for the x independent of z; i.e., depend only on x and y. With e e ˆ e components of E and H which depend only on the z this in mind, we further decompose components of H 3 e e components of E and H. Here they are: as follows: ! ex = e ez ez H hx (x, y)e−jkz z (6.124) ∂E ∂H −j e + ωµ (6.136) Ex = 2 +kz kρ ∂x ∂y ey = e hy (x, y)e−jkz z (6.125) H ! ez = e ez ez H hz (x, y)e−jkz z (6.126) ∂ E +j ∂ H ey = −kz (6.137) E + ωµ kρ2 ∂y ∂x e e e where hx (x, y), hy (x, y), and hz (x, y) represent the ! ez ez remaining factors. One advantage of this +j ∂E ∂H e Hx = 2 ωǫ − kz (6.138) decomposition is that partial derivatives with respect kρ ∂y ∂x to z reduce to algebraic operations; i.e.: ! ez ez −j ∂ E ∂ H ey = H ωǫ + kz (6.139) ex ∂H e x (6.127) kρ2 ∂x ∂y = −jkz e hx (x, y)e−jkz z = −jkz H ∂z ey ∂H e y (6.128) where = −jkz e hy (x, y)e−jkz z = −jkz H kρ2 , β 2 − kz2 (6.140) ∂z e ∂ Hz e z (6.129) Why define a parameter called “kρ ”? Note from the = −jkz e hz (x, y)e−jkz z = −jkz H ∂z definition that β 2 = kρ2 + kz2 . Further, note that this equation is an expression of the Pythagorean theorem, We now substitute Equations 6.124–6.126 into which relates the lengths of the sides of a right Equations 6.121–6.123 and then use Equations 6.127–6.129 to eliminate partial derivatives triangle. Since β is the “overall” phase propagation constant, and kz is the phase propagation constant for with respect to z. This yields: ˆ direction in the waveguide, kρ propagation in the z ! ez 1 ∂ H must be associated with variation in fields in ex = ey + jkz H (6.130) E ˆ. In the cylindrical directions perpendicular to z jωǫ ∂y ! coordinate system, this is the ρˆ direction, hence the e ey = 1 e x − ∂ Hz E −jkz H (6.131) subscript “ρ.” We shall see later that kρ plays a special jωǫ ∂x 3 Students are encouraged to derive these for themselves – only ! ex , begin with Equae e algebra is required. Tip to get started: To get E ∂ Hy ∂ Hx ez = 1 e y using Equation 6.134, and then solve for E − (6.132) tion 6.130, eliminate H jωǫ ∂x ∂y e find:

!

Ex .

CHAPTER 6. WAVEGUIDES

110 role in determining the structure of fields within the waveguide, and this provides additional motivation to identify this quantity explicitly in the field equations. Summarizing: If you know the wave is unidirectional, e and H e in the then knowledge of the components of E direction of propagation is sufficient to determine e and H. e There each of the remaining components of E is one catch, however: For this to yield sensible results, kρ may not be zero. So, ironically, these expressions don’t work in the case of a uniform plane ez = H ez = 0 wave, since such a wave would have E and kz = β (so kρ = 0), yielding values of zero divided by zero for each remaining field component. The same issue arises for any other transverse electromagnetic (TEM) wave. For non-TEM waves – and, in particular, unidirectional waves in waveguides ez or H e z must be non-zero and kρ will be – either E non-zero. In this case, Equations 6.136–6.139 are both usable and useful since they allow determination of all field components given just the z components. In fact, we may further exploit this simplicity by taking one additional step: Decomposition of the unidirectional wave into transverse electric (TE) and transverse magnetic (TM) components. In this case, ez = 0, and TM means simply TE means simply that E e z = 0. For a wave which is either TE or TM, that H Equations 6.136–6.139 reduce to one term each.

6.8 Rectangular Waveguide: TM Modes [m0223]

A rectangular waveguide is a conducting cylinder of rectangular cross section used to guide the propagation of waves. Rectangular waveguide is commonly used for the transport of radio frequency signals at frequencies in the SHF band (3–30 GHz) and higher. The fields in a rectangular waveguide consist of a number of propagating modes which depends on the electrical dimensions of the waveguide. These modes are broadly classified as either transverse magnetic (TM) or transverse electric (TE). In this section, we consider the TM modes. Figure 6.6 shows the geometry of interest. Here the walls are located at x = 0, x = a, y = 0, and y = b; thus, the cross-sectional dimensions of the waveguide are a and b. The interior of the waveguide is presumed to consist of a lossless material exhibiting real-valued permeability µ and real-valued permittivity ǫ, and the walls are assumed to be perfectly-conducting. Let us limit our attention to a region within the waveguide which is free of sources. Expressed in phasor form, the electric field intensity within the waveguide is governed by the wave equation

where

e + β2E e =0 ∇2 E √ β = ω µǫ

(6.141) (6.142)

Figure 6.6: Geometry for analysis of fields in a rectangular waveguide.

6.8. RECTANGULAR WAVEGUIDE: TM MODES Equation 6.141 is a partial differential equation. This equation, combined with boundary conditions imposed by the perfectly-conducting plates, is sufficient to determine a unique solution. This solution is most easily determined in Cartesian coordinates, as we shall now demonstrate. First we e in Cartesian coordinates: express E e =x ex + y ey + z ez ˆE ˆE ˆE E

111 e z = 0; i.e., is transverse (perpendicular) to the H direction of propagation. Thus, the TM component is ez . completely determined by E Equations 6.136–6.139 simplify to become: e ex = −j kz ∂ Ez E 2 kρ ∂x e ey = −j kz ∂ Ez E 2 kρ ∂y

(6.143)

This facilitates the decomposition of Equation 6.141 ˆ, y ˆ , and z ˆ into separate equations governing the x e components of E: ex + β 2 E ex = 0 ∇2 E ey + β 2 E ey = 0 ∇2 E ez + β 2 E ez = 0 ∇2 E

Next we observe that the operator ∇2 may be expressed in Cartesian coordinates as follows:

e e x = +j ωµ ∂ Ez H 2 kρ ∂y

(6.144) (6.145) (6.146) where

e e y = −j ωµ ∂ Ez H kρ2 ∂x kρ2 , β 2 − kz2

(6.151) (6.152) (6.153) (6.154)

(6.155)

and kz is the phase propagation constant; i.e., the wave is assumed to propagate according to e−jkz z .

ez , which Now let us address the problem of finding E (6.147) will then completely determine the TM field. As in ez can be represented Section 6.7, we recognize that E so the equations governing the Cartesian components as the propagation factor e−jkz z times a factor that e may be written as follows: of E describes variation with respect to the remaining spatial dimensions x and y: 2 2 2 ∂ e ∂ e ∂ e 2e E + E + E + β E = 0 (6.148) x x x x ∂x2 ∂y 2 ∂z 2 ez = eez (x, y)e−jkz z E (6.156) 2 2 2 ∂ e ∂ e ∂ e ey = 0 (6.149) Ey + 2 Ey + 2 Ey + β 2 E Substitution of this expression into Equation 6.150 ∂x2 ∂y ∂z 2 2 2 and dividing out the common factor of e−jkz z yields: ∂ e ∂ e ∂ e 2e E + E + E + β E = 0 (6.150) z z z z ∂x2 ∂y 2 ∂z 2 ∂2 ∂2 eez + 2 eez − kz2 eez + β 2 eez = 0 (6.157) 2 ∂x ∂y In general, we expect the total field in the waveguide The last two terms may be combined using to consist of unidirectional waves propagating in the Equation 6.155, yielding: +ˆ z and −ˆ z directions. We may analyze either of these waves; then the other wave is easily derived via ∂2 ∂2 symmetry, and the total field is simply a linear eez + 2 eez + kρ2 eez = 0 (6.158) 2 ∂x ∂y combination (superposition) of these waves. With this in mind, we limit our focus to the wave propagating This is a partial differential equation for eez in the in the +ˆ z direction. variables x and y. This equation may be solved using the technique of separation of variables. In this In Section 6.7, it is shown that all components of the technique, we recognize that eez (x, y) can be written electric and magnetic fields can be easily calculated e e as the product of a function X(x) which depends only once Ez and Hz are known. The problem is further on x, and a function Y (y) that depends only on y. simplified by decomposing the unidirectional wave That is, into TM and TE components. In this decomposition, the TM component is defined by the property that eez (x, y) = X(x)Y (y) (6.159) ∇2 =

∂2 ∂2 ∂2 + + ∂x2 ∂y 2 ∂z 2

CHAPTER 6. WAVEGUIDES

112 Substituting this expression into Equation 6.158, we obtain: Y

∂2 ∂2 X + X Y + kρ2 XY = 0 ∂x2 ∂y 2

(6.160)

Next dividing through by XY , we obtain: 1 ∂2 1 ∂2 X + Y + kρ2 = 0 X ∂x2 Y ∂y 2

(6.161)

Note that the first term depends only on x, the second term depends only on y, and the remaining term is a constant. Therefore, the sum of the first and second terms is a constant; namely −kρ2 . Since these terms depend on either x or y, and not both, the first term must equal some constant, the second term must equal some constant, and these constants must sum to −kρ2 . Therefore, we are justified in separating the equation into two equations as follows:

The solution is essentially complete except for the values of the constants A, B, C, D, kx , and ky . The values of these constants are determined by applying the relevant electromagnetic boundary condition. In e that is this case, it is required that the component of E tangent to a perfectly-conducting wall must be zero. e is tangent to all four ˆ component of E Note that the z walls; therefore: ez (x = 0) = 0 E ez (x = a) = 0 E

(6.162) (6.163)

where the new constants kx2 and ky2 must satisfy kx2 + ky2 = kρ2

(6.164)

Now multiplying Equations 6.162 and 6.163 by X and Y , respectively, we find: 2

∂ X + kx2 X = 0 ∂x2 ∂2 Y + ky2 Y = 0 ∂y 2

(6.165) (6.166)

(6.167) (6.168)

where A, B, C, and D – like kx and ky – are constants to be determined. At this point, we observe that the wave we seek can be expressed as follows: ez = eez (x, y)e E

−jkz z

(6.172)

X (x = 0) = 0

(6.174)

X (x = a) = 0 Y (y = 0) = 0 Y (y = b) = 0

(6.175) (6.176) (6.177)

(6.173)

Evaluating these conditions using Equations 6.167 and 6.168 yields: A·1+B·0=0 A cos (kx a) + B sin (kx a) = 0

(6.178) (6.179)

C ·1+D·0=0 C cos (ky b) + D sin (ky b) = 0

(6.180) (6.181)

Equations 6.178 and 6.180 can be satisfied only if A = 0 and C = 0, respectively. Subsequently, Equations 6.179 and 6.181 reduce to:

These are familiar one-dimensional differential equations. The solutions are:4 X = A cos (kx x) + B sin (kx x) Y = C cos (ky y) + D sin (ky y)

(6.171)

ez (y = 0) = 0 E ez (y = b) = 0 E

Referring to Equation 6.169, these boundary conditions in turn require:

2

1 ∂ X + kx2 = 0 X ∂x2 1 ∂2 Y + ky2 = 0 Y ∂y 2

(6.170)

sin (kx a) = 0

(6.182)

sin (ky b) = 0

(6.183)

This in turn requires: mπ , m = 0, 1, 2... a nπ ky = , n = 0, 1, 2... b

kx =

(6.184) (6.185)

Each positive integer value of m and n leads to a ez known as a mode. Solutions (6.169) valid expression for E for which m = 0 or n = 0 yield kx = 0 or ky = 0, 4 Students are encouraged to confirm that these are correct by confirming that they are solutions to Equations 6.165 and 6.165, re- respectively. These correspond to zero-valued fields, spectively. and therefore are not of interest. The most general = X(x) Y (y) e−jkz z

6.9. RECTANGULAR WAVEGUIDE: TE MODES ez must account for all non-trivial expression for E modes. Summarizing: ez = E

where

∞ X ∞ X

m=1 n=1

ez(m,n) E

113

6.9 Rectangular Waveguide: TE Modes

(6.186) [m0225]

A rectangular waveguide is a conducting cylinder of rectangular cross section used to guide the propagation of waves. Rectangular waveguide is commonly used for the transport of radio frequency signals at frequencies in the SHF band (3–30 GHz) and higher. The fields in a rectangular waveguide consist of a number of propagating modes which depends on the electrical dimensions of the waveguide. These modes are broadly classified as (6.188) either transverse magnetic (TM) or transverse electric (TE). In this section, we consider the TE modes.

    ez(m,n) , E (m,n) sin mπ x sin nπ y e−jkz(m,n) z E 0 a b (6.187) (m,n) where E0 is an arbitrary constant (consolidating the constants B and D), and, since kx2 + ky2 = kρ2 , β 2 − kz2 : kz(m,n)

=

r

ω 2 µǫ −

 mπ 2 a



 nπ 2 b

Summarizing: e z = 0) component of the unidiThe TM (H rectional (+ˆ z-traveling) wave in a rectangular waveguide is completely determined by Equation 6.186, and consists of modes as defined by Equations 6.187, 6.188, 6.184, and 6.185. The remaining non-zero field components can be determined using Equations 6.151–6.154. It is customary and convenient to refer to the TM modes in a rectangular waveguide using the notation “TMmn .” For example, the mode TM12 is given by Equation 6.187 with m = 1 and n = 2. (m,n)

Finally, note that values of kz obtained from Equation 6.188 are not necessarily real-valued. It is (m,n) apparent that for any given value of m, kz will be imaginary-valued for all values of n greater than some value. Similarly, it is apparent that for any given value (m,n) of n, kz will be imaginary-valued for all values of m greater than some value. This phenomenon is common to both TM and TE components, and so is addressed in a separate section (Section 6.10).

Figure 6.7 shows the geometry of interest. Here the walls are located at x = 0, x = a, y = 0, and y = b; thus, the cross-sectional dimensions of the waveguide are a and b. The interior of the waveguide is presumed to consist of a lossless material exhibiting real-valued permeability µ and real-valued permittivity ǫ, and the walls are assumed to be perfectly-conducting. Let us limit our attention to a region within the waveguide which is free of sources. Expressed in phasor form, the magnetic field intensity within the waveguide is governed by the wave equation:

where

e + β2H e =0 ∇2 H √ β = ω µǫ

(6.189) (6.190)

Additional Reading: • “Waveguide (radio frequency)” on Wikipedia. • “Separation of variables” on Wikipedia.

Figure 6.7: Geometry for analysis of fields in a rectangular waveguide.

CHAPTER 6. WAVEGUIDES

114 Equation 6.189 is a partial differential equation. This equation, combined with boundary conditions imposed by the perfectly-conducting plates, is sufficient to determine a unique solution. This solution is most easily determined in Cartesian coordinates, as we shall now demonstrate. First we e in Cartesian coordinates: express H

ez = 0; i.e., is transverse (perpendicular) to the E direction of propagation. Thus, the TE component is ez . completely determined by H Equations 6.136–6.139 simplify to become:

This facilitates the decomposition of Equation 6.189 ˆ, y ˆ , and z ˆ into separate equations governing the x e components of H:

e e x = −j kz ∂ Hz H kρ2 ∂x

e =x ex + y ey + z ez ˆH ˆH ˆH H

ex + β2H ex = 0 ∇2 H ey + β2H ey = 0 ∇2 H ez + β2H ez = 0 ∇2 H

Next we observe that the operator ∇2 may be expressed in Cartesian coordinates as follows:

e ex = −j ωµ ∂ Hz E 2 kρ ∂y e ey = +j ωµ ∂ Hz E 2 kρ ∂x

(6.191)

(6.192) (6.193) (6.194) where

e e y = −j kz ∂ Hz H kρ2 ∂y kρ2 , β 2 − kz2

(6.199) (6.200) (6.201) (6.202)

(6.203)

and kz is the phase propagation constant; i.e., the wave is assumed to propagate according to e−jkz z .

e z , which Now let us address the problem of finding H (6.195) will then completely determine the TE field. As in e z can be represented Section 6.7, we recognize that H so the equations governing the Cartesian components as the propagation factor e−jkz z times a factor that e may be written as follows: of H describes variation with respect to the remaining spatial dimensions x and y: ∂2 e ∂2 e ∂2 e 2 e Hx + 2 Hx + 2 Hx + β Hx = 0 (6.196) ∂x2 ∂y ∂z ez = e hz (x, y)e−jkz z (6.204) H 2 2 2 ∂ e ∂ e ∂ e e y = 0 (6.197) Hy + 2 Hy + 2 Hy + β 2 H Substitution of this expression into Equation 6.198 ∂x2 ∂y ∂z and dividing out the common factor of e−jkz z yields: 2 2 2 ∂ e ∂ e ∂ e 2 e H + H + H + β H = 0 (6.198) z z z z ∂x2 ∂y 2 ∂z 2 ∂2 e ∂2 e h + hz − kz2 e hz + β 2 e hz = 0 (6.205) z ∂x2 ∂y 2 In general, we expect the total field in the waveguide The last two terms may be combined using to consist of unidirectional waves propagating in the Equation 6.203, yielding: +ˆ z and −ˆ z directions. We may analyze either of these waves; then the other wave is easily derived via ∂2 e ∂2 e symmetry, and the total field is simply a linear h + hz + kρ2 e hz = 0 (6.206) z ∂x2 ∂y 2 combination (superposition) of these waves. With this in mind, we limit our focus to the wave propagating This is a partial differential equation for e hz in the in the +ˆ z direction. variables x and y. This equation may be solved using the technique of separation of variables. In this In Section 6.7, it is shown that all components of the technique, we recognize that e hz (x, y) can be written electric and magnetic fields can be easily calculated e e as the product of a function X(x) which depends only once Ez and Hz are known. The problem is further on x, and a function Y (y) that depends only on y. simplified by decomposing the unidirectional wave That is, into TM and TE components. In this decomposition, e the TE component is defined by the property that hz (x, y) = X(x)Y (y) (6.207) ∇2 =

∂2 ∂2 ∂2 + + ∂x2 ∂y 2 ∂z 2

6.9. RECTANGULAR WAVEGUIDE: TE MODES

115

The solution is essentially complete except for the values of the constants A, B, C, D, kx , and ky . The values of these constants are determined by applying (6.208) the relevant electromagnetic boundary condition. In e that this case, it is required that any component of E is tangent to a perfectly-conducting wall must be zero. Therefore: ey (x = 0) = 0 (6.209) E (6.218)

Substituting this expression into Equation 6.206, we obtain: Y

∂2 ∂2 X + X Y + kρ2 XY = 0 ∂x2 ∂y 2

Next dividing through by XY , we obtain: 1 ∂2 1 ∂2 X + Y + kρ2 = 0 X ∂x2 Y ∂y 2

Note that the first term depends only on x, the second term depends only on y, and the remaining term is a constant. Therefore, the sum of the first and second terms is a constant; namely −kρ2 . Since these terms depend on either x or y, and not both, the first term must equal some constant, the second term must equal some constant, and these constants must sum to −kρ2 . Therefore, we are justified in separating the equation into two equations as follows: 1 ∂2 X + kx2 = 0 X ∂x2 1 ∂2 Y + ky2 = 0 Y ∂y 2 where the new constants

kx2

and

ky2

(6.212)

(6.213)

X = A cos (kx x) + B sin (kx x)

(6.215)

Y = C cos (ky y) + D sin (ky y)

(6.216)

where A, B, C, and D – like kx and ky – are constants to be determined. At this point, we observe that the wave we seek can be expressed as follows:

= X(x) Y (y) e−jkz z

5 Students

(6.220) (6.221)

∂ X (x = 0) = 0 ∂x ∂ X (x = a) = 0 ∂x ∂ Y (y = 0) = 0 ∂y ∂ Y (y = b) = 0 ∂y

(6.222) (6.223) (6.224) (6.225)

Evaluating the partial derivatives and dividing out common factors of kx and ky , we find: −A sin(kx · 0) + B cos(kx · 0) = 0 −A sin(kx · a) + B cos(kx · a) = 0

−C sin(ky · 0) + D cos(ky · 0) = 0 −C sin(ky · b) + D cos(ky · b) = 0

(6.226) (6.227) (6.228) (6.229)

Evaluating: (6.214)

These are familiar one-dimensional differential equations. The solutions are:5

ez = e H hz (x, y)e−jkz z

(6.219)

Referring to Equation 6.217 and employing Equations 6.199–6.202, we obtain:

(6.211)

Now multiplying Equations 6.210 and 6.211 by X and Y , respectively, we find: ∂2 X + kx2 X = 0 ∂x2 ∂2 Y + ky2 Y = 0 ∂y 2

ex (y = b) = 0 E

(6.210)

must satisfy

kx2 + ky2 = kρ2

ey (x = a) = 0 E ex (y = 0) = 0 E

(6.217)

are encouraged to confirm that these are correct by confirming that they are solutions to Equations 6.213 and 6.213, respectively.

−A · 0 + B · 1 = 0 −A sin (kx a) + B cos (kx a) = 0 −C · 0 + D · 1 = 0 −C sin (ky b) + D cos (ky b) = 0

(6.230) (6.231) (6.232) (6.233)

Equations 6.230 and 6.232 can be satisfied only if B = 0 and D = 0, respectively. Subsequently, Equations 6.231 and 6.233 reduce to: sin (kx a) = 0 sin (ky b) = 0 This in turn requires: mπ , m = 0, 1, 2... kx = a nπ ky = , n = 0, 1, 2... b

(6.234) (6.235)

(6.236) (6.237)

CHAPTER 6. WAVEGUIDES

116

Each positive integer value of m and n leads to a valid imaginary-valued for all values of n greater than some e z known as a mode. Summarizing: expression for H value. Similarly, it is apparent that for any given value (m,n) of n, kz will be imaginary-valued for all values of ∞ X ∞ X m greater than some value. This phenomenon is (m,n) ez = e H H (6.238) z common to both TE and TM components, and so is m=0 n=0 addressed in a separate section (Section 6.10). where Additional Reading:     e (m,n) , H (m,n) cos mπ x cos nπ y e−jkz(m,n) z H z 0 a b (6.239) • “Waveguide (radio frequency)” on Wikipedia. (m,n) where H0 is an arbitrary constant (consolidating • “Separation of variables” on Wikipedia. the constants A and C), and, since kx2 + ky2 = kρ2 , β 2 − kz2 : kz(m,n)

=

r

ω 2 µǫ −

 mπ 2 a



 nπ 2 b

(6.240)

Summarizing: ez = 0) component of the unidiThe TE (E rectional (+ˆ z-traveling) wave in a rectangular waveguide is completely determined by Equation 6.238, and consists of modes as defined by Equations 6.239, 6.240, 6.236, and 6.237. The remaining non-zero field components can be determined using Equations 6.199–6.202. It is customary and convenient to refer to the TE modes in a rectangular waveguide using the notation “TEmn .” For example, the mode TE12 is given by Equation 6.239 with m = 1 and n = 2. Although Equation 6.238 implies the existence of a TE00 mode, it should be noted that this wave has no non-zero electric field components. This can be determined mathematically by following the procedure outlined above. However, this is also ex is constant for the readily confirmed as follows: E TE00 mode because kρ = 0 for this mode, however, ex must be zero to meet the boundary conditions on E ey is the walls at y = 0 and y = b. Similarly, E e constant for the TE00 mode, however, Ey must be zero to meet the boundary conditions on the walls at x = 0 and x = b. (m,n)

Finally, note that values of kz obtained from Equation 6.240 are not necessarily real-valued. It is (m,n) apparent that for any given value of m, kz will be

6.10. RECTANGULAR WAVEGUIDE: PROPAGATION CHARACTERISTICS

6.10 Rectangular Waveguide: Propagation Characteristics

The “+” sign option corresponds to a wave that grows exponentially in magnitude with increasing z, which is non-physical behavior. Therefore: (m,n)

e−jkz

[m0224]

In this section, we consider the propagation characteristics of TE and TM modes in rectangular waveguides. Because these modes exhibit the same phase dependence on z, findings of this section apply equally to both sets of modes. Recall that the TM modes in a rectangular waveguide are given by:     ez(m,n) = E (m,n) sin mπ x sin nπ y e−jkz(m,n) z E 0 a b (6.241) (m,n) where E0 is an arbitrary constant (determined in part by sources), and: r  mπ 2  nπ 2 (m,n) kz = ω 2 µǫ − − (6.242) a b

The TE modes in a rectangular waveguide are:     e z(m,n) = H (m,n) cos mπ x cos nπ y e−jkz(m,n) z H 0 a b (6.243) (m,n) where H0 is an arbitrary constant (determined in part by sources). Cutoff frequency. First, observe that values of (m,n) kz obtained from Equation 6.242 are not necessarily real-valued. For any given value of m, (m,n) 2 (kz ) will be negative for all values of n greater than some value. Similarly, for any given value of n, (m,n) 2 (kz ) will be negative for all values of m greater than some value. Should either of these conditions occur, we find:  2  mπ 2  nπ 2 kz(m,n) = ω 2 µǫ − −

a b

Since ω = 2πf :

r

mπ 2  nπ 2 + (6.249) 2π µǫ a b r  1 m 2  n 2 = √ + (6.250) 2 µǫ a b √ Note that 1/ µǫ is the phase velocity vp for the medium used in the waveguide. With this in mind, let us define: 1 vpu , √ (6.251) µǫ f>

1 √

This is the phase velocity in an unbounded medium having the same permeability and permittivity as the interior of the waveguide. Thus: r vpu  m 2  n 2 f> + (6.252) 2 a b

In other words, the mode (m, n) avoids being cut off if the frequency is high enough to meet this criterion. Thus, it is useful to make the following definition: 2 r = −α (6.244) vpu  m 2  n 2 fmn , + (6.253) 2 a b where α is a positive real-valued constant. So: kz(m,n) = ±jα

(6.245) The cutoff frequency fmn (Equation 6.253) is the lowest frequency for which the mode (m, n) is able to propagate (i.e., not cut off).

Subsequently: (m,n)

e−jkz

z

= e−j(±jα)z = e±αz

(6.246)

CHAPTER 6. WAVEGUIDES

118 Example 6.4. Cutoff frequencies for WR-90. WR-90 is a popular implementation of rectangular waveguide. WR-90 is air-filled with dimensions a = 22.86 mm and b = 10.16 mm. Determine cutoff frequencies and, in particular, the lowest frequency at which WR-90 can be used. Solution. Since WR-90 is air-filled, µ ≈ µ0 , √ ǫ ≈ ǫ0 , and vpu ≈ 1 µ0 ǫ0 ∼ = 3.00 × 108 m/s. Cutoff frequencies are given by Equation 6.253. Recall that there are no non-zero TE or TM modes with m = 0 and n = 0. Since a > b, the lowest non-zero cutoff frequency is achieved when m = 1 and n = 0. In this case, Equation 6.253 yields f10 = 6.557 GHz; this is the lowest frequency that is able to propagate efficiently in the waveguide. The next lowest cutoff frequency is f20 = 13.114 GHz. The third lowest cutoff frequency is f01 = 14.754 GHz. The lowest-order TM mode that is non-zero and not cut off is TM11 (f11 = 16.145 GHz). Phase velocity. The phase velocity for a wave propagating within a rectangular waveguide is greater than that of electromagnetic radiation in unbounded space. For example, the phase velocity of any propagating mode in a vacuum-filled waveguide is greater than c, the speed of light in free space. This is a surprising result. Let us first derive this result and then attempt to make sense of it. Phase velocity vp in the rectangular waveguide is given by vp ,

ω

(6.254)

(m,n)

kz

vpu vp = q 2 1 − (fmn /f )

(6.257)

For any propagating mode, f > fmn ; subsequently, vp > vpu . In particular, vp > c for a vacuum-filled waveguide. How can this not be a violation of fundamental physics? As noted in Section 6.1, phase velocity is not necessarily the speed at which information travels, but is merely the speed at which a point of constant phase travels. To send information, we must create a disturbance in the otherwise sinusoidal excitation presumed in the analysis so far. The complex field structure creates points of constant phase that travel faster than the disturbance is able to convey information, so there is no violation of physical principles. Group velocity. As noted in Section 6.1, the speed at which information travels is given by the group velocity vg . In unbounded space, vg = vp , so the speed of information is equal to the phase velocity in that case. In a rectangular waveguide, the situation is different. We find: ! (m,n) −1 ∂kz vg = ∂ω q 2 = vpu 1 − (fmn /f )

(6.258) (6.259)

which is always less than vpu for a propagating mode.

ω

=q

√ Note that 1/ µǫ is vpu , as defined earlier. Employing Equation 6.253 and also noting that ω = 2πf , Equation 6.256 may be rewritten in the following form:

2

2

(6.255)

Note that group velocity in the waveguide depends on frequency in two ways. First, because fmn takes on Immediately we observe that phase velocity seems to different values for different modes, group velocity is different for different modes. Specifically, be different for different modes. Dividing the √ numerator and denominator by β = ω µǫ, we obtain: higher-order modes propagate more slowly than lower-order modes having the same frequency. This is 1 1 known as modal dispersion. Secondly, note that the vp = √ r i h µǫ group velocity of any given mode depends on 2 2 −1 2 1 − (ω µǫ) (mπ/a) + (nπ/b) frequency. This is known as chromatic dispersion. (6.256) ω 2 µǫ − (mπ/a) − (nπ/b)

6.10. RECTANGULAR WAVEGUIDE: PROPAGATION CHARACTERISTICS The speed of a signal within a rectangular waveguide is given by the group velocity of the associated mode (Equation 6.259). This speed is less than the speed of propagation in unbounded media having the same permittivity and permeability. Speed depends on the ratio fmn /f , and generally decreases with increasing frequency for any given mode.

Example 6.5. Speed of propagating in WR-90. Revisiting WR-90 waveguide from Example 6.4: What is the speed of propagation for a narrowband signal at 10 GHz? Solution. Let us assume that “narrowband” here means that the bandwidth is negligible relative to the center frequency, so that we need only consider the center frequency. As previously determined, the lowest-order propagating mode is TE10 , for which f10 = 6.557 GHz. The next-lowest cutoff frequency is f20 = 13.114 GHz. Therefore, only the TE10 mode is available for this signal. The group velocity for this mode at the frequency of interest is given by Equation 6.259. Using this equation, the speed of propagation is found to be ∼ = 2.26 × 108 m/s, which is about 75.5% of c.

[m0212]

119

120

CHAPTER 6. WAVEGUIDES

Image Credits c Sevenchw (C. Wang), Fig. 6.1: https://commons.wikimedia.org/wiki/File:Geometry for analysis of fields in parallel plate waveguide.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), modified. c Sevenchw (C. Wang), Fig. 6.2: https://commons.wikimedia.org/wiki/File:TE component of electric field in parallel plate waveguide.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 6.3: https://commons.wikimedia.org/wiki/File:Magnitude of the two lowest-order TE modes in parallel plate waveguide.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 6.4: https://commons.wikimedia.org/wiki/File:TM component of electric field in parallel plate waveguide.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 6.5: https://commons.wikimedia.org/wiki/File:TM component of electric field in parallel plate waveguide.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), modified.

Chapter 7

Transmission Lines Redux 7.1 Parallel Wire Transmission Line [m0188]

A parallel wire transmission line consists of wires separated by a dielectric spacer. Figure 7.1 shows a common implementation, commonly known as “twin lead.” The wires in twin lead line are held in place by a mechanical spacer comprised of the same low-loss dielectric material that forms the jacket of each wire. Very little of the total energy associated with the electric and magnetic fields lies inside this material, so the jacket and spacer can usually be neglected for the purposes of analysis and electrical design. Parallel wire transmission line is often employed in radio applications up to about 100 MHz as an alternative to coaxial line. Parallel wire line has the advantages of lower cost and lower loss than coaxial line in this frequency range. However, parallel wire line lacks the self-shielding property of coaxial cable; i.e., the electromagnetic fields of coaxial line are isolated by the outer conductor, whereas those of

c C. Wang CC BY SA 4.0

Figure 7.2: Parallel wire transmission line structure and design parameters. parallel wire line are exposed and prone to interaction with nearby structures and devices. This prevents the use of parallel wire line in many applications. Another discriminator between parallel wire line and coaxial line is that parallel wire line is differential.1 The conductor geometry is symmetric and neither conductor is favored as a signal datum (“ground”). Thus, parallel wire line is commonly used in applications where the signal sources and/or loads are also differential; common examples are the dipole antenna and differential amplifiers.2 Figure 7.2 shows a cross-section of parallel wire line. Relevant parameters include the wire diameter, d; and the center-to-center spacing, D.

c SpinningSpark, Inductiveload CC BY SA 3.0 (modified)

Figure 7.1: Twin lead, a commonly-encountered form of parallel wire transmission line.

1 The references in “Additional Reading” at the end of this section

may be helpful if you are not familiar with this concept. 2 This is in contrast to “single-ended” line such as coaxial line, which has conductors of different cross-sections and the outer conductor is favored as the datum.

c 2020 S.W. Ellingson CC BY SA 4.0. https://doi.org/10.21061/electromagnetics-vol-2 Electromagnetics Vol. 2.

CHAPTER 7. TRANSMISSION LINES REDUX

122

Because the wire separation D is typically much greater than the wire diameter d, D/d ≫ 1 and so q 2 (D/d) − 1 ≈ D/d. This leads to the simplified expressions L′ ≈

µ0 ln (2D/d) (D ≫ d) π

(7.5)

πǫ (D ≫ d) ln (2D/d)

(7.6)

C′ ≈

Now returning to Equation 7.2: r 1 µ0 Z0 ≈ ln (2D/d) (7.7) Figure 7.3: Structure of the electric and magnetic π ǫ fields for a cross-section of parallel wire line. In this p case, the wave is propagating away from the viewer. Noting that ǫ = ǫr ǫ0 and µ0 /ǫ0 , η0 , we obtain c S. Lally CC BY SA 4.0

The associated field structure is transverse electromagnetic (TEM) and is therefore completely described by a single cross-section along the line, as shown in Figure 7.3. Expressions for these fields exist, but are complex and not particularly useful except as a means to calculate other parameters of interest. One of these parameters is, of course, the characteristic impedance since this parameter plays an important role in the analysis and design of systems employing transmission lines. The characteristic impedance may be determined using the “lumped element” transmission line model using the following expression: s R′ + jωL′ (7.1) Z0 = G′ + jωC ′

Z0 ≈

1 η0 √ ln (2D/d) π ǫr

(7.8)

The characteristic impedance of parallel wire line, assuming low-loss conditions and wire spacing much greater than wire diameter, is given by Equation 7.8. Observe that the characteristic impedance of parallel wire line increases with increasing D/d. Since this ratio is large, the characteristic impedance of parallel wire line tends to be large relative to common values of other kinds of TEM transmission line, such as coaxial line and microstrip line. An example follows.

where R′ , G′ , C ′ , and L′ are the resistance, Example 7.1. 300 Ω twin-lead. conductance, capacitance, and inductance per unit length, respectively. This analysis is considerably A commonly-encountered form of parallel wire simplified by neglecting loss; therefore, let us assume transmission line is 300 Ω twin-lead. Although the “low-loss” conditions R′ ≪ ωL′ and G′ ≪ ωC ′ . implementations vary, the wire diameter is Then we find: r usually about 1 mm and and the wire spacing is ′ L usually about 6 mm. The relative permittivity of Z0 ≈ (low loss) (7.2) C′ the medium ǫr ≈ 1 for the purposes of and the problem is reduced to determining inductance calculating transmission line parameters, since and capacitance of the transmission line. These are the jacket and spacer have only a small effect on   q the fields. For these values, Equation 7.8 gives µ0 2 L′ = ln (D/d) + (D/d) − 1 (7.3) Z0 ≈ 298 Ω, as expected. π πǫ  C′ =  (7.4) q Under the assumption that the wire jacket/spacer 2 ln (D/d) + (D/d) − 1 material has a negligible effect on the electromagnetic

7.2. MICROSTRIP LINE REDUX fields, and that the line is suspended in air so that ǫr ≈ 1, the phase velocity vp for a parallel wire line is approximately that of any electromagnetic wave in free space; i.e., c. In practical twin-lead, the effect of a plastic jacket/spacer material is to reduce the phase velocity by a few percent up to about 20%, depending on the materials and details of construction. So in practice vp ≈ 0.8c to 0.9c for twin-lead line. Additional Reading: • “Twin-lead” on Wikipedia. • “Differential signaling” on Wikipedia. • Sec. 8.7 (“Differential Circuits”) in S.W. Ellingson, Radio Systems Engineering, Cambridge Univ. Press, 2016.

123

7.2 Microstrip Line Redux [m0186]

A microstrip transmission line consists of a narrow metallic trace separated from a metallic ground plane by a slab of dielectric material, as shown in Figure 7.4. This is a natural way to implement a transmission line on a printed circuit board, and so accounts for an important and expansive range of applications. The reader should be aware that microstrip is distinct from stripline, which is a distinct type of transmission line; see “Additional Reading” at the end of this section for disambiguation of these terms. A microstrip line is single-ended3 in the sense that the conductor geometry is asymmetric and the one conductor – namely, the ground plane – also normally serves as ground for the source and load. The spacer material is typically a low-loss dielectric material having permeability approximately equal to that of free space (µ ≈ µ0 ) and relative permittivity ǫr in the range of 2 to about 10 or so. The structure of a microstrip line is similar to that of a parallel plate waveguide (Section 6.6), with the obvious difference that one of the “plates” has finite length in the direction perpendicular to the direction of propagation. Despite this difference, the parallel plate waveguide provides some useful insight into the operation of the microstrip line. Microstrip line is nearly always operated below the cutoff frequency of 3 The reference in “Additional Reading” at the end of this section may be helpful if you are not familiar with this concept.

t

0

r

l

W

h

c 7head7metal7 CC BY SA 3.0 (modified)

Figure 7.4: Microstrip transmission line structure and design parameters.

124

CHAPTER 7. TRANSMISSION LINES REDUX

with propagating waves lies directly underneath the trace, and Figure 7.5 provides a relatively accurate impression of the fields in this region. The characteristic impedance Z0 may be determined using the “lumped element” transmission line model using the following expression: s R′ + jωL′ (7.9) Z0 = Figure 7.5: Approximate structure of the electric and G′ + jωC ′ magnetic fields within microstrip line, assuming TM0 ′ ′ ′ ′ operation. The fields outside the line are possibly sig- where R , G , C , and L are the resistance, nificant, complicated, and not shown. In this case, the conductance, capacitance, and inductance per unit length, respectively. For the present analysis, nothing wave is propagating away from the viewer. is lost by assuming loss is negligible; therefore, let us assume R′ ≪ ωL′ and G′ ≪ ωG′ , yielding all but the TM0 mode, whose cutoff frequency is zero. r L′ This guarantees that only the TM0 mode can (7.10) Z0 ≈ propagate. The TM0 mode has the form of a uniform C′ plane wave, as shown in Figure 7.5. The electric field Thus the problem is reduced to determining is oriented in the direction perpendicular to the plates, capacitance and inductance of the transmission line. and magnetic field is oriented in the direction parallel to the plates. The direction of propagation E × H for Wide microstrip line resembles a parallel-plate the TM0 mode always points in the same direction; capacitor whose plate spacing d is very small namely, along the axis of the transmission line. compared to the plate area A. In this case, fringing Therefore, microstrip lines nominally exhibit fields may be considered negligible and one finds that transverse electromagnetic (TEM) field structure. the capacitance C is given by The limited width W of the trace results in a “fringing field” – i.e., significant deviations from TM0 field structure in the dielectric beyond the edges of the trace and above the trace. The fringing fields may play a significant role in determining the characteristic impedance Z0 . Since Z0 is an important parameter in the analysis and design of systems using transmission lines, we are motivated to not only determine Z0 for the microstrip line, but also to understand how variation in W affects Z0 . Let us address this issue by considering the following three special cases, in order: • W ≫ h, which we shall refer to as the “wide” case. • W ≪ h, which we shall refer to as the “narrow” case. • W ∼ h; i.e., the intermediate case in which h and W equal to within an order of magnitude or so. Wide case. If W ≫ h, most of the energy associated

C≈

ǫA (parallel plate capacitor) d

(7.11)

In terms of wide microstrip line, A = W l where l is length, and d = h. Therefore: C′ ,

C ǫW ≈ l h

(W ≫ h)

(7.12)

To determine L′ , consider the view of the microstrip line shown in Figure 7.6. Here a current source applies a steady current I on the left, and a resistive load closes the current loop on the right. Ampere’s law for magnetostatics and the associated “right hand rule” require that H is directed into the page and is approximately uniform. The magnetic flux between the trace and the ground plane is Z Φ= B · ds (7.13) S

where S is the surface bounded by the current loop and B = µ0 H. In the present problem, the area of S is hl and H is approximately constant over S, so: Φ ≈ µ0 Hhl

(7.14)

7.2. MICROSTRIP LINE REDUX

125

c C. Wang CC BY SA 4.0

c S. Lally CC BY SA 4.0

Figure 7.6: View from the side of a microstrip line, Figure 7.7: Electric and magnetic fields lines for narused to determine L′ . row (W ≪ h) microstrip line. where H is the magnitude of H. Next, recall that: L,

Φ I

(7.15)

So, we may determine L if we are able to obtain an expression for I in terms of H. This can be done as follows. First, note that we can express I in terms of the current density on the trace; i.e., I = Js W

(7.16)

where Js is the current density (SI base units of A/m) on the trace. Next, we note that the boundary condition for H on the trace requires that the discontinuity in the tangent component of H going from the trace (where the magnetic field intensity is H) to beyond the trace (i.e., outside the transmission line, where the magnetic field intensity is approximately zero) be equal to the surface current. Thus, Js ≈ H and I = Js W ≈ HW

(7.17)

Returning to Equation 7.15, we find: L,

Φ µ0 Hhl µ0 hl ≈ = I HW W

(7.18)

L µ0 h ≈ (W ≫ h) l W

(7.19)

Subsequently, L′ ,

Now the characteristic impedance is found to be s r r L′ µ0 h/W µ0 h Z0 ≈ = (7.20) ≈ ′ C ǫW/h ǫ W p The factor µ0 /ǫ is recognized as the wave impedance η. It is convenient to express this in terms

of the free space wave impedance η0 , leading to the expression we seek: η0 h Z0 ≈ √ ǫr W

p µ0 /ǫ0 ,

(W ≫ h)

(7.21)

The characteristic impedance of “wide” (W ≫ h) microstrip line is given approximately by Equation 7.21. It is worth noting that the characteristic impedance of wide transmission line is proportional to h/W . The factor h/W figures prominently in the narrow- and intermediate-width cases as well, as we shall soon see. Narrow case. Figure 7.5 does not accurately depict the fields in the case that W ≪ h. Instead, much of the energy associated with the electric and magnetic fields lies beyond and above the trace. Although these fields are relatively complex, we can develop a rudimentary model of the field structure by considering the relevant boundary conditions. In particular, the tangential component of E must be zero on the surfaces of the trace and ground plane. Also, we expect magnetic field lines to be perpendicular to the electric field lines so that the direction of propagation is along the axis of the line. These insights lead to Figure 7.7. Note that the field lines shown in Figure 7.7 are quite similar to those for parallel wire line (Section 7.1). If we choose diameter d = W and wire spacing D = 2h, then the fields of the microstrip line in the dielectric spacer must be similar to those of the parallel wire line between the lines. This is shown in

126

CHAPTER 7. TRANSMISSION LINES REDUX Figure 7.8. Note that the fields above the dielectric spacer in the microstrip line will typically be somewhat different (since the material above the trace is typically different; e.g., air). However, it remains true that we expect most of the energy to be contained in the dielectric, so we shall assume that this difference has a relatively small effect. Next, we continue to refine the model by introducing the ground plane, as shown in Figure 7.9: The fields in the upper half-space are not perturbed when we do this, since the fields in Figure 7.8 already satisfy the boundary conditions imposed by the new ground plane. Thus, we see that the parallel wire transmission line provides a pretty good guide to the structure of the fields for the narrow transmission line, at least in the dielectric region of the upper half-space.

We are now readyp to estimate the characteristic impedance Z0 ≈ L′ /C ′ of low-loss narrow Figure 7.8: Noting the similarity of the fields in narmicrostrip line. Neglecting the differences noted row microstrip line to those in a parallel wire line. above, we estimate that this is roughly equal to the characteristic impedance of the analogous (d = W and D = 2h) parallel wire line. Under this assumption, we obtain from Equation 7.8: Z0 ∼

1 η0 √ ln (4h/W ) (W ≪ h) π ǫr

(7.22)

This estimate will exhibit only “order of magnitude” accuracy (hence the “∼” symbol), since the contribution from the fields in half of the relevant space is ignored. However we expect that Equation 7.22 will accurately capture the dependence of Z0 on the parameters ǫr and h/W . These properties can be useful for making fine adjustments to a microstrip design. The characteristic impedance of “narrow” (W ≪ h) microstrip line can be roughly approximated by Equation 7.22. Figure 7.9: Modeling the fields in narrow microstrip line as those of a parallel wire line, now introducing Intermediate case. An expression for Z0 in the intermediate case – even a rough estimate – is difficult the ground plane. to derive, and beyond the scope of this book. However, we are not completely in the dark, since we can reasonably anticipate that Z0 in the intermediate case should be intermediate to estimates provided by the wide and narrow cases. This is best demonstrated by example. Figure 7.10 shows estimates of Z0 using

7.2. MICROSTRIP LINE REDUX

127 for which W ′ ≈ W . Although complicated, this formula should not be completely surprising. For example, note the parameters h and W appear in this formula as the factor 4h/W , which we encountered in the “narrow” approximation. Also, we see that the formula indicates Z0 increases with increasing h/W and decreasing ǫr , as we determined in both the “narrow” and “wide” cases. The following example provides a demonstration for the very common case of microstrip implementation in FR4 circuit board material. Example 7.2. Characteristic impedance of microstrip lines in FR4.

Figure 7.10: Z0 for FR4 as a function of h/W , as determined by the “wide” and “narrow” approximations, along with the Wheeler 1977 formula. Note that the vertical and horizontal axes of this plot are in log scale. the wide and narrow expressions (blue and green curves, respectively) for a particular implementation of microstrip line (see Example 7.2 for details). Since Z0 varies smoothly with h/W , it is reasonable to expect that simply averaging the values generated using the wide and narrow assumptions would give a pretty good estimate when h/W ∼ 1. However, it is also possible to obtain an accurate estimate directly. A widely-accepted and broadly-applicable formula is provided in Wheeler 1977 (cited in “Additional Reading” at the end of this section). This expression is valid over the full range of h/W , not merely the intermediate case of h/W ∼ 1. Here it is: 42.4 Ω × Z0 ≈ √ ǫ +1 " r !# r 4h 1 + 1/ǫ r ln 1 + ′ K + K 2 + π2 W 2

FR4 printed circuit boards consist of a substrate having h ∼ = 1.575 mm and ǫr ≈ 4.5. Figure 7.10 shows Z0 as a function of h/W , as determined by the wide and narrow approximations, along with the value obtained using the Wheeler 1977 formula. The left side of the plot represents the wide condition, whereas the right side of this plot represents the narrow condition. The Wheeler 1977 formula is an accurate estimate over the entire horizontal span. Note that the wide approximation is accurate only for h/W < 0.1 or so, improving with decreasing h/W as expected. The narrow approximation overestimates Z0 by a factor of about 1.4 (i.e., about 40%). This is consistent with our previous observation that this approximation should yield only about “order of magnitude” accuracy. Nevertheless, the narrow approximation exhibits approximately the same rate of increase with h/W as does the Wheeler 1977 formula. Also worth noting from Figure 7.10 is the important and commonly-used result that Z0 ≈ 50 Ω is obtained for h/W ≈ 0.5. Thus, a 50 Ω microstrip line in FR4 has a trace width of about 3 mm.

(7.23) where K,

14 + 8/ǫr 11



4h W′



(7.24)

and W ′ is W adjusted to account for the thickness t of the microstrip line. Typically t ≪ W and t ≪ h,

A useful “take away” from this example is that the wide and narrow approximations serve as useful guides for understanding how Z0 changes as a function of h/W and ǫr . This is useful especially for making adjustments to a microstrip line design once an accurate value of Z0 is obtained from some other

CHAPTER 7. TRANSMISSION LINES REDUX

128

method; e.g., using the Wheeler 1977 formula or from Next, ǫr,ef f is crudely approximated as the average of the relative permittivity of the dielectric slab and the measurements. relative permittivity of free space; i.e.,: FR4 circuit board construction is so common that the ǫr + 1 result from the previous example deserves to be ǫr,ef f ≈ (7.27) 2 highlighted: In FR4 printed circuit board construction (substrate thickness 1.575 mm, relative permittivity ≈ 4.5, negligible trace thickness), Z0 ≈ 50 Ω requires a trace width of about 3 mm. Z0 scales roughly in proportion to h/W around this value.

Various refinements exist to improve on this approximation; however, in practice, variations in the value of ǫr for the dielectric due to manufacturing processes typically make a more precise estimate irrelevant. Using this concept, we obtain

Simpler approximations for Z0 are also commonly employed in the design and analysis of microstrip lines. These expressions are limited in the range of h/W for which they are valid, and can usually be shown to be special cases or approximations of Equation 7.23. Nevertheless, they are sometimes useful for quick “back of the envelope” calculations.

λ=

λ0 2π 2π = √ =√ β β0 ǫr,ef f ǫr,ef f

(7.28)

where λ0 is the free-space wavelength c/f . Similarly the phase velocity vp can be estimated using the relationship ω c vp = = √ (7.29) β ǫr,ef f

Wavelength in microstrip line. An accurate general i.e., the phase velocity in microstrip is slower than c formula for wavelength λ in microstrip line is √ by a factor of ǫr,ef f . similarly difficult to derive. A useful approximate technique employs a result from the theory of uniform plane waves in unbounded media. For such waves, Example 7.3. Wavelength and phase velocity in the phase propagation constant β is given by microstrip in FR4 printed circuit boards. √ β = ω µǫ (7.25) FR4 is a low-loss fiberglass epoxy dielectric that is commonly used to make printed circuit boards It turns out that the electromagnetic field structure for (see “Additional Reading” at the end of this the guided wave in a microstrip line is similar in the section). For FR4, ǫr ≈ 4.5. The effective sense that it exhibits TEM field structure, as does the relative permittivity is therefore: uniform plane wave. However, the guided wave is different in that it exists in both the dielectric spacer ǫr,ef f ≈ (4.5 + 1)/2 = 2.75 and the air above the spacer. Thus, we presume that β Thus, we estimate the phase velocity for for the guided wave can be approximated as that of a √the wave guided by this line to be about c/ 2.75; uniform plane wave in unbounded media having the i.e., 60% of c. Similarly, the wavelength of this same permeability µ0 but a different relative wave is about 60% of the free space wavelength. permittivity, which we shall assign the symbol ǫr,ef f In practice, these values are found to be slightly (for “effective relative permittivity”). Then: less; typically 50%–55%. The difference is √ attributable to the crude approximation of β ≈ ω µ0 ǫr,ef f ǫ0 (low-loss microstrip) √ Equation 7.27. =β ǫ (7.26) 0

r,ef f

In other words, the phase propagation constant in a microstrip line can be approximated as the free-space √ phase propagation β0 , ω µ0 ǫ0 times a correction √ factor ǫr,ef f .

Additional Reading: • “Microstrip” on Wikipedia.

7.3. ATTENUATION IN COAXIAL CABLE • “Printed circuit board” on Wikipedia.

129

7.3 Attenuation in Coaxial Cable

• “Stripline” on Wikipedia. • “Single-ended signaling” on Wikipedia.

[m0189]

• Sec. 8.7 (“Differential Circuits”) in S.W. Ellingson, Radio Systems Engineering, Cambridge Univ. Press, 2016.

In this section, we consider the issue of attenuation in coaxial transmission line. Recall that attenuation can be interpreted in the context of the “lumped element” • H.A. Wheeler, “Transmission Line Properties of equivalent circuit transmission line model as the contributions of the resistance per unit length R′ and a Strip on a Dielectric Sheet on a Plane,” IEEE ′ ′ Trans. Microwave Theory & Techniques, Vol. 25, conductance per unit length G . In this model, R represents the physical resistance in the inner and No. 8, Aug 1977, pp. 631–47. outer conductors, whereas G′ represents loss due to • “FR-4” on Wikipedia. current flowing directly between the conductors through the spacer material. The parameters used to describe the relevant features of coaxial cable are shown in Figure 7.11. In this figure, a and b are the radii of the inner and outer conductors, respectively. σic and σoc are the conductivities (SI base units of S/m) of the inner and outer conductors, respectively. Conductors are assumed to be non-magnetic; i.e., having permeability µ equal to the free space value µ0 . The spacer material is assumed to be a lossy dielectric having relative permittivity ǫr and conductivity σs . Resistance per unit length. The resistance per unit length is the sum of the resistances of the inner and outer conductor per unit length. The resistance per unit length of the inner conductor is determined by σic and the effective cross-sectional area through which the current flows. The latter is equal to the circumference 2πa times the skin depth δic of the

b a σic σoc

ϵr σs

Figure 7.11: Parameters defining the design of a coaxial cable.

130

CHAPTER 7. TRANSMISSION LINES REDUX

A simpler way to deal with these two cases is to represent them both using the single expression 1 ′   for δic ≪ a (7.30) Ric ≈ 1 1 C (2πa · δic ) σic ′ R ≈ + (7.38) 2πδic σic a b This expression is only valid for δic ≪ a because otherwise the cross-sectional area through which the where C = 0 in Case I and C = 1 in Case II. current flows is not well-modeled as a thin ring near Conductance per unit length. The conductance per the surface of the conductor. Similarly, we find the unit length of coaxial cable is simply that of the resistance per unit length of the outer conductor is associated coaxial structure at DC; i.e., 1 ′ Roc ≈ for δoc ≪ t (7.31) 2πσs (2πb · δoc ) σoc (7.39) G′ = ln (b/a) where δoc is the skin depth of the outer conductor and Unlike resistance, the conductance is independent of t is the thickness of the outer conductor. Therefore, frequency, at least to the extent that σs is independent the total resistance per unit length is of frequency. ′ ′ R′ = Ric + Roc Attenuation. The attenuation of voltage and current 1 1 (7.32) waves as they propagate along the cable is + ≈ (2πa · δic ) σic (2πb · δoc ) σoc represented by the factor e−αz , where z is distance traversed along the cable. It is possible to find an Recall that skin depth depends on conductivity. expression for α in terms of the material and Specifically: geometry parameters using: p p (7.33) δic = 2/ωµσic p γ , (R′ + jωL′ ) (G′ + jωC ′ ) = α + jβ (7.40) δoc = 2/ωµσoc (7.34) where L′ and C ′ are the inductance per unit length Expanding Equation 7.32 to show explicitly the and capacitance per unit length, respectively. These dependence on conductivity, we find: are given by µ   L′ = ln (b/a) (7.41) 1 1 1 ′ 2π p (7.35) R ≈ + √ √ b σoc 2π 2/ωµ0 a σic and 2πǫ0 ǫr (7.42) C′ = At this point it is convenient to identify two particular ln (b/a) cases for the design of the cable. In the first case, In principle we could solve Equation 7.40 for α. “Case I,” we assume σoc ≫ σic . Since b > a, we However, this course of action is quite tedious, and a have in this case simpler approximate approach facilitates some   additional insights. In this approach, we define 1 1 p R′ ≈ √ parameters αR associated with R′ and αG associated a σ 2π 2/ωµ0 ic with G′ such that 1 1 = (Case I) (7.36) 2πδic σic a e−αR z e−αG z = e−(αR +αG )z = e−αz (7.43) inner conductor, so:

In the second case, “Case II,” we assume σoc = σic . which indicates In this case, we have   α = αR + αG 1 1 1 ′ p R ≈ + √ √ b σic 2π 2/ωµ0 a σic Next we postulate   1 1 1 R′ (7.37) = + (Case II) α R ≈ KR 2πδic σic a b Z0

(7.44)

(7.45)

7.3. ATTENUATION IN COAXIAL CABLE

131

where Z0 is the characteristic impedance Z0 ≈

η0 1 b (low loss) √ ln 2π ǫr a

(7.46)

and where KR is a unitless constant to be determined. The justification for Equation 7.45 is as follows: First, αR must increase monotonically with increasing R′ . Second, R′ must be divided by an impedance in order to obtain the correct units of 1/m. Using similar reasoning, we postulate αG ≈ KG G′ Z0

(7.47)

where KG is a unitless constant to be determined. The following example demonstrates the validity of Equations 7.45 and 7.47, and will reveal the values of KR and KG . Example 7.4. Attenuation constant for RG-59. RG-59 is a popular form of coaxial cable having the parameters a ∼ = 0.292 mm, b ∼ = 1.855 mm, 7 ∼ σic = 2.28 × 10 S/m, σs ∼ = 5.9 × 10−5 S/m, and ǫr ∼ = 2.25. The conductivity σoc of the outer conductor is difficult to quantify because it consists of a braid of thin metal strands. However, σoc ≫ σic , so we may assume Case I; i.e., σoc ≫ σic , and subsequently C = 0. Figure 7.12 shows the components αG and αR computed for the particular choice KR = KG = 1/2. The figure also shows αG + αR , along with α computed using Equation 7.40. We find that the agreement between these values is very good, which is compelling evidence that the ansatz is valid and KR = KG = 1/2.

Figure 7.12: Comparison of α = Re {γ} to αR , αG , and αR + αG for KR = KG = 1/2. The result for α has been multiplied by 1.01; otherwise the curves would be too close to tell apart. may consider αR and αG independently. Let us first consider αG : 1 ′ G Z0 2 1 2πσs 1 η0 ≈ · · √ ln (b/a) 2 ln (b/a) 2π ǫr η0 σs = (7.48) √ 2 ǫr

αG ,

It is clear from this result that αG is minimized by √ minimizing σs / ǫr . Interestingly the physical dimensions a and b have no discernible effect on αG . Now we consider αR : R′ 2Z0 1 (1/2πδic σic ) [1/a + C/b] = √  2 (1/2π) η0 / ǫr ln (b/a) √ ǫr [1/a + C/b] · (7.49) = 2η0 δic σic ln (b/a) p Now making the substitution δic = 2/ωµ0 σic in order to make the dependences on the constitutive parameters explicit, we find: r 1 ωµ0 ǫr [1/a + C/b] αR = √ · (7.50) σic ln (b/a) 2 2 · η0 αR ,

Note that there is nothing to indicate that the results demonstrated in the example are not generally true. Thus, we come to the following conclusion: The attenuation constant α ≈ αG + αR where αG , R′ /2Z0 and αR , G′ Z0 /2. Minimizing attenuation. Let us now consider if there are design choices which minimize the attenuation of coaxial cable. Since α = αR + αG , we

CHAPTER 7. TRANSMISSION LINES REDUX

132 Here we see that αR is minimized by minimizing ǫr /σic . It’s not surprising to see that we should maximize σic . However, it’s a little surprising that we should minimize ǫr . Furthermore, this is in contrast to αG , which is minimized by maximizing ǫr . Clearly there is a tradeoff to be made here. To determine the parameters of this tradeoff, first note that the result depends on frequency: Since αR dominates over αG at sufficiently high frequency (as demonstrated in Figure 7.12), it seems we should minimize ǫr if the intended frequency of operation is sufficiently high; otherwise the optimum value is frequency-dependent. However, σs may vary as a function of ǫr , so a general conclusion about optimum values of σs and ǫr is not appropriate. However, we also see that αR – unlike αG – depends on a and b. This implies the existence of a generally-optimum geometry. To find this geometry, we minimize αR by taking the derivative with respect to a, setting the result equal to zero, and solving for a and/or b. Here we go: r 1 ωµ0 ǫr ∂ [1/a + C/b] ∂ αR = √ · ∂a σic ∂a ln (b/a) 2 2 · η0 (7.51) This derivative is worked out in an addendum at the end of this section. Using the result from the addendum, the right side of Equation 7.51 can be written as follows:   r ωµ0 ǫr 1/a + C/b 1 −1 √ · 2 + σic a ln (b/a) a ln2 (b/a) 2 2 · η0 (7.52) In order for ∂αR /∂a = 0, the factor in the square brackets above must be equal to zero. After a few steps of algebra, we find: ln (b/a) = 1 +

C b/a

somewhere between 2.72 and 3.59, with the precise value depending on the relative conductivity of the inner and outer conductors. Substituting these values of b/a into Equation 7.46, we obtain: 59.9 Ω 76.6 Ω Z0 ≈ √ to √ ǫr ǫr

as the range of impedances of coaxial cable corresponding to physical designs that minimize attenuation. Equation 7.56 gives the range of characteristic impedances that minimize attenuation for coaxial transmission lines. The precise value within this range depends on the ratio of the conductivity of the outer conductor to that of the inner conductor.

Since ǫr ≥ 1, the impedance that minimizes attenuation is less for dielectric-filled cables than it is for air-filled cables. For example, let us once again consider the RG-59 from Example 7.4. In that case, ǫr ∼ = 2.25 and C = 0, indicating Z0 ≈ 39.9 Ω is optimum for attenuation. The actual characteristic impedance of Z0 is about 75 Ω, so clearly RG-59 is not optimized for attenuation. This is simply because other considerations apply, including power handling capability (addressed in Section 7.4) and the convenience of standard values (addressed in Section 7.5).

Addendum: Derivative of a2 ln(b/a). Evaluation of Equation 7.51 requires finding the derivative of (7.53) a2 ln(b/a) with respect to a. Using the chain rule, we find:

In Case I (σoc ≫ σic ), C = 0 so: b/a = e ∼ = 2.72 (Case I)

(7.56)

(7.54)

In Case II (σoc = σic ), C = 1. The resulting equation can be solved by plotting the function, or by a few iterations of trial and error; either way one quickly finds b/a ∼ (7.55) Note = 3.59 (Case II) Summarizing, we have found that α is minimized by choosing the ratio of the outer and inner radii to be

       ∂ b ∂ 2 b a2 ln = a ln ∂a a ∂a a    ∂ b +a2 ln ∂a a

∂ 2 a = 2a ∂a

(7.57)

(7.58)

7.4. POWER HANDLING CAPABILITY OF COAXIAL CABLE

7.4 Power Handling Capability of Coaxial Cable

and   ∂ b ∂ ln = [ln (b) − ln (a)] ∂a a ∂a ∂ =− ln (a) ∂a 1 =− a

[m0190]

(7.59)

So:        1 ∂ b b a2 ln = [2a] ln + a2 − ∂a a a a   b −a (7.60) = 2a ln a This result is substituted for a2 ln(b/a) in Equation 7.51 to obtain Equation 7.52.

133

The term “power handling” refers to maximum power that can be safely transferred by a transmission line. This power is limited because when the electric field becomes too large, dielectric breakdown and arcing may occur. This may result in damage to the line and connected devices, and so must be avoided. Let Epk be the maximum safe value of the electric field intensity within the line, and let Pmax be the power that is being transferred under this condition. This section addresses the following question: How does one design a coaxial cable to maximize Pmax for a given Epk ? We begin by finding the electric potential V within the cable. This can be done using Laplace’s equation: ∇2 V = 0

(7.61)

Using the cylindrical (ρ, φ, z) coordinate system with the z axis along the inner conductor, we have ∂V /∂φ = 0 due to symmetry. Also we set ∂V /∂z = 0 since the result should not depend on z. Thus, we have:   ∂V 1 ∂ ρ =0 (7.62) ρ ∂ρ ∂ρ Solving for V , we have V (ρ) = A ln ρ + B

(7.63)

where A and B are arbitrary constants, presumably determined by boundary conditions. Let us assume a voltage V0 measured from the inner conductor (serving as the “+” terminal) to the outer conductor (serving as the “−” terminal). For this choice, we have: V (a) = V0 V (b) = 0

→ →

A ln a + B = V0

(7.64)

A ln b + B = 0

(7.65)

Subtracting the second equation from the first and solving for A, we find A = −V0 / ln (b/a). Subsequently, B is found to be V0 ln (b) / ln (b/a), and so V (ρ) =

−V0 V0 ln (b) ln ρ + ln (b/a) ln (b/a)

(7.66)

CHAPTER 7. TRANSMISSION LINES REDUX

134 The electric field intensity is given by: E = −∇V Again we have ∂V /∂φ = ∂V /∂z = 0, so

for optimum power handling. In other words, 1.65 is the ratio of the radii of the outer and inner conductors (7.67) that maximizes the power that can be safely handled by the cable.

Equation 7.75 suggests that ǫr should be maximized (7.68) in order to maximize power handling, and you wouldn’t be wrong for noting that, however, there are (7.69) some other factors that may indicate otherwise. For example, a material with higher ǫr may also have (7.70) higher σs , which means more current flowing through the spacer and thus more ohmic heating. This problem is so severe that cables that handle high RF Note that the maximum electric field intensity in the spacer occurs at ρ = a; i.e., at the surface of the inner power often use air as the spacer, even though it has the lowest possible value of ǫr . Also worth noting is conductor. Therefore: that σic and σoc do not matter according to the V0 Epk = (7.71) analysis we’ve just done; however, to the extent that a ln (b/a) limited conductivity results in significant ohmic heating in the conductors – which we have also not The power transferred by the line is maximized when considered – there may be something to consider. the impedances of the source and load are matched to Suffice it to say, the actionable finding here concerns the ratio of the radii; the other parameters have not Z0 . In this case, the power transferred is V02 /2Z0 . Recall that the characteristic impedance Z0 is given in been suitably constrained by this analysis. the “low-loss” case as √   Substituting e for b/a in Equation 7.72, we find: 1 η0 b Z0 ≈ (7.72) √ ln 2π ǫr a 30.0 Ω Z0 ≈ √ (7.78) Therefore, the maximum safe power is ǫr ∂ V ∂ρ   −V0 V0 ln (b) ∂ ln ρ + = −ˆ ρ ∂ρ ln (b/a) ln (b/a) V0 = +ˆ ρ ρ ln (b/a)

E = −ˆ ρ

Pmax =

V02 2Z0

(7.73)

2 Epk a2 ln2 (b/a) ≈ √  2 · (1/2π) η0 / ǫr ln (b/a)

=

2 πEpk √ a2 ln (b/a) η 0 / ǫr

This is the characteristic impedance of coaxial line that optimizes power handling, subject to the caveats (7.74) identified above. For air-filled cables, we obtain 30 Ω. Since ǫr ≥ 1, this optimum impedance is less for (7.75) dielectric-filled cables than it is for air-filled cables.

Now let us consider if there is a value of a which maximizes Pmax . We do this by seeing if ∂Pmax /∂a = 0 for some values of a and b. The derivative is worked out in an addendum at the end of this section. Using the result from the addendum, we find: 2 πEpk ∂ Pmax = √ [2a ln (b/a) − a] ∂a η 0 / ǫr

(7.76)

For the above expression to be zero, it must be true that 2 ln (b/a) − 1 = 0. Solving for b/a, we obtain: √ b = e∼ = 1.65 a

Summarizing: The power handling capability of coaxial transmission line is optimized when the ratio of radii of the outer to inner conductors b/a is about 1.65. For the air-filled cables typically used in highpower applications, this corresponds to a characteristic impedance of about 30 Ω.

Addendum: Derivative of (1/a + C/b) / ln(b/a). Evaluation of Equation 7.75 requires finding the (7.77) derivative of (1/a + C/b) / ln(b/a) with respect to a.

7.5. WHY 50 OHMS?

135

Using the chain rule, we find:        b ∂ 1/a + C/b ∂ 1 C −1 = + ln ∂a ln (b/a) ∂a a b a     1 C ∂ −1 b + + ln a b ∂a a (7.79) Note

∂ ∂a



1 C + a b



=−

1 a2

(7.80)

To handle the quantity in the second set of square brackets, first define v = ln u, where u = b/a. Then:     ∂ −1 ∂u ∂ −1 ∂v v = v ∂a ∂v ∂u ∂a      1  −ba−2 = −v −2 u    h i  b a  = − ln−2 −ba−2 a b   1 b = ln−2 (7.81) a a So:       ∂ 1/a + C/b 1 b = − 2 ln−1 ∂a ln (b/a) a a     1 C 1 −2 b + + ln a b a a (7.82) This result is substituted in Equation 7.75 to obtain Equation 7.76.

7.5 Why 50 Ohms? [m0191]

The quantity 50 Ω appears in a broad range of applications across the field of electrical engineering. In particular, it is a very popular value for the characteristic impedance of transmission line, and is commonly specified as the port impedance for signal sources, amplifiers, filters, antennas, and other RF components. So, what’s special about 50 Ω? The short answer is “nothing.” In fact, other standard impedances are in common use – prominent among these is 75 Ω. It is shown in this section that a broad range of impedances – on the order of 10s of ohms – emerge as useful values based on technical considerations such as minimizing attenuation, maximizing power handling, and compatibility with common types of antennas. Characteristic impedances up to 300 Ω and beyond are useful in particular applications. However, it is not practical or efficient to manufacture and sell products for every possible impedance in this range. Instead, engineers have settled on 50 Ω as a round number that lies near the middle of this range, and have chosen a few other values to accommodate the smaller number of applications where there may be specific compelling considerations. So, the question becomes “what makes characteristic impedances in the range of 10s of ohms particularly useful?” One consideration is attenuation in coaxial cable. Coaxial cable is by far the most popular type of transmission line for connecting devices on separate printed circuit boards or in separate enclosures. The attenuation of coaxial cable is addressed in Section 7.3. In that section, it is shown that attenuation is minimized for characteristic √ √ impedances in the range (60 Ω) / ǫr to (77 Ω) / ǫr , where ǫr is the relative permittivity of the spacer material. So, we find that Z0 in the range 60 Ω to 77 Ω is optimum for air-filled cable, but more like 40 Ω to 50 Ω for cables using a plastic spacer material having typical ǫr ≈ 2.25. Thus, 50 Ω is clearly a reasonable choice if a single standard value is to be established for all such cable. Coaxial cables are often required to carry high power signals. In such applications, power handling

136 capability is also important, and is addressed in Section 7.4. In that section, we find the power handling capability of coaxial cable is optimized when the ratio of radii of the outer to inner conductors b/a is about 1.65. For the air-filled cables typically used in high-power applications, this corresponds to a characteristic impedance of about 30 Ω. This is significantly less than the 60 Ω to 77 Ω that minimizes attenuation in air-filled cables. So, 50 Ω can be viewed as a compromise between minimizing attenuation and maximizing power handling in air-filled coaxial cables. Although the preceding arguments justify 50 Ω as a standard value, one can also see how one might make a case for 75 Ω as a secondary standard value, especially for applications where attenuation is the primary consideration. Values of 50 Ω and 75 Ω also offer some convenience when connecting RF devices to antennas. For example, 75 Ω is very close to the impedance of the commonly-encountered half-wave dipole antenna (about 73 + j42 Ω), which may make impedance matching to that antenna easier. Another commonly-encountered antenna is the quarter-wave monopole, which exhibits an impedance of about 36 + j21 Ω, which is close to 50 Ω. In fact, we see that if we desire a single characteristic impedance that is equally convenient for applications involving either type of antenna, then 50 Ω is a reasonable choice. A third commonly-encountered antenna is the folded half-wave dipole. This type of antenna is similar to a half-wave dipole but has better bandwidth, and is commonly used in FM and TV systems and land mobile radio (LMR) base stations. A folded half-wave dipole has an impedance of about 300 Ω and is balanced (not single-ended); thus, there is a market for balanced transmission line having Z0 = 300 Ω. However, it is very easy and inexpensive to implement a balun (a device which converts the dipole output from balanced to unbalanced) while simultaneously stepping down impedance by a factor of 4; i.e., to 75 Ω. Thus, we have an additional application for 75 Ω coaxial line. Finally, note that it is quite simple to implement microstrip transmission line having characteristic impedance in the range 30 Ω to 75 Ω. For example,

CHAPTER 7. TRANSMISSION LINES REDUX 50 Ω on commonly-used 1.575 mm FR4 requires a width-to-height ratio of about 2, so the trace is about 3 mm wide. This is a very manageable size and easily implemented in printed circuit board designs. Additional Reading: • “Dipole antenna” on Wikipedia. • “Monopole antenna” on Wikipedia. • “Balun” on Wikipedia. [m0187]

7.5. WHY 50 OHMS?

137

Image Credits c SpinningSpark, Inductiveload, Fig. 7.1: https://commons.wikimedia.org/wiki/File:Twin-lead cable dimension.svg, CC BY-SA 3.0 (https://creativecommons.org/licenses/by-sa/3.0/). Minor modifications. c Sevenchw (C. Wang), Fig. 7.2: https://commons.wikimedia.org/wiki/File:Parallel wire transmission line structure and design parameters.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), Fig. 7.3: https://commons.wikimedia.org/wiki/File:Electric and Magnetic Field of Wire Cross-Section.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c 7head7metal7, https://commons.wikimedia.org/wiki/File:Microstrip scheme.svg, Fig. 7.4: CC BY-SA 3.0 (https://creativecommons.org/licenses/by-sa/3.0/). Minor modifications from the original. c Sevenchw (C. Wang), Fig. 7.6: https://commons.wikimedia.org/wiki/File:View from the side of a microstrip line.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), Fig. 7.7: https://commons.wikimedia.org/wiki/File:Electric and Magnetic Fields for Microstrip.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/).

Chapter 8

Optical Fiber 8.1 Optical Fiber: Method of Operation [m0178]

In its simplest form, optical fiber consists of concentric regions of dielectric material as shown in Figure 8.1. A cross-section through the fiber reveals a circular region of transparent dielectric material through which light propagates. This is surrounded by a jacket of dielectric material commonly referred to as cladding. A characteristic of the design of any optical fiber is that the permittivity of the fiber is greater than the permittivity of the cladding. As explained in Section 5.11, this creates conditions necessary for total internal reflection. The mechanism of total internal reflection contains the light within the fiber. In the discipline of optics, the permittivity of a

c S. Lally CC BY-SA 4.0

material is commonly quantified in terms of its index of refraction. Index of refraction is the square root of relative permittivity, and is usually assigned the symbol n. Thus, if we define the relative permittivities ǫr,f , ǫf /ǫ0 for the fiber and ǫr,c , ǫc /ǫ0 for the cladding, then nf , nc ,





ǫr,f

(8.1)

ǫr,c

(8.2)

and nf > nc

(8.3)

Figure 8.2 illustrates total internal reflection in an optical fiber. In this case, a ray of light in the fiber is incident on the boundary with the cladding. We may treat the light ray as a uniform plane wave. To see why, consider that optical wavelengths range from 120 nm to 700 nm in free space. Wavelength is slightly shorter than this in fiber; specifically, by a factor equal to the square root of the relative permittivity. The fiber is on the order of millimeters in diameter, which is about 4 orders of magnitude greater than the wavelength. Thus, from the perspective of the light ray, the fiber appears to be an unbounded half-space sharing a planar boundary with the cladding, which also appears to be an unbounded half-space. Continuing under this presumption, the criterion for total internal reflection is (from Section 5.11): r nc ǫr,c = arcsin (8.4) θi ≥ arcsin ǫr,f nf

Figure 8.1: Construction of the simplest form of opti- As long as rays of light approach from angles that cal fiber. satisfy this criterion, the associated power remains in

c 2020 S.W. Ellingson CC BY SA 4.0. https://doi.org/10.21061/electromagnetics-vol-2 Electromagnetics Vol. 2.

8.1. OPTICAL FIBER: METHOD OF OPERATION

139 Additional Reading: • “Optical fiber” on Wikipedia.

c S. Lally CC BY-SA 4.0

Figure 8.2: Total internal reflection in optical fiber. the fiber, and is reflected onward. Otherwise, power is lost into the cladding. Example 8.1. Critical angle for optical fiber. Typical values of nf and nc for an optical fiber are 1.52 and 1.49, respectively. What internal angle of incidence is required to maintain total internal reflection? Solution. Using Equation 8.4, θi must be greater than about 78.8◦ . In practice this is quite reasonable since we desire light to be traveling approximately parallel the axis of the fiber (θi ≈ 90◦ ) anyway. The total internal reflection criterion imposes a limit on the radius of curvature of fiber optic cable. If fiber optic cable is bent such that the radius of curvature is too small, the critical angle will be exceeded at the bend. This will occur even for light rays which are traveling perfectly parallel to the axis of the fiber before they arrive at the bend. Note that the cladding serves at least two roles. First, it determines the critical angle for total internal reflection, and subsequently determines the minimum radius of curvature for lossless operation of the fiber. Second, it provides a place for the evanescent surface waves associated with total internal reflection (Section 5.12) to exist without interference from objects in contact with the fiber.

CHAPTER 8. OPTICAL FIBER

140

8.2 Acceptance Angle

Now invoking a trigonometric identity:

[m0192]

In this section, we consider the problem of injecting light into a fiber optic cable. The problem is illustrated in Figure 8.3. In this figure, we see light incident from a medium having index of refraction n0 , with angle of incidence θi . The light is transmitted with angle of transmission θ2 into the fiber, and is subsequently incident on the surface of the cladding with angle of incidence θ3 . For light to propagate without loss within the cable, it is required that nc sin θ3 ≥ nf

(8.5)

since this criterion must be met in order for total internal reflection to occur. Now consider the constraint that Equation 8.5 imposes on θi . First, we note that θ3 is related to θ2 as follows: π θ3 = − θ2 (8.6) 2 therefore  π − θ2 (8.7) sin θ3 = sin 2 = cos θ2 (8.8) so cos θ2 ≥

nc nf

(8.9)

n2c n2f

(8.10)

Squaring both sides, we find: cos2 θ2 ≥

1 − sin2 θ2 ≥

n2c n2f

(8.11)

sin2 θ2 ≤ 1 −

n2c n2f

(8.12)

so:

Now we relate the θ2 to θi using Snell’s law: sin θ2 =

n0 sin θi nf

(8.13)

so Equation 8.12 may be written: n2 n20 sin2 θi ≤ 1 − 2c 2 nf nf Now solving for sin θi , we obtain: 1q 2 sin θi ≤ nf − n2c n0

(8.14)

(8.15)

This result indicates the range of angles of incidence which result in total internal reflection within the fiber. The maximum value of θi which satisfies this condition is known as the acceptance angle θa , so:  q  1 θa , arcsin n2f − n2c (8.16) n0 This leads to the following insight: In order to effectively launch light in the fiber, it is necessary for the light to arrive from within a cone having half-angle θa with respect to the axis of the fiber. The associated cone of acceptance is illustrated in Figure 8.4. It is also common to define the quantity numerical aperture NA as follows: 1q 2 NA , nf − n2c (8.17) n0

c S. Lally CC BY-SA 4.0

Figure 8.3: Injecting light into a fiber optic cable.

Note that n0 is typically very close to 1 (corresponding to incidence from air),q so it is common to see NA defined as simply n2f − n2c . This parameter is commonly used in lieu of the acceptance angle in datasheets for fiber optic cable.

8.3. DISPERSION IN OPTICAL FIBER Example 8.2. Acceptance angle. Typical values of nf and nc for an optical fiber are 1.52 and 1.49, respectively. What are the numerical aperture and the acceptance angle? Solution. Using Equation 8.17 and presuming n0 = 1, we find NA ∼ = 0.30. Since sin θa = NA, we find θa = 17.5◦ . Light must arrive from within 17.5◦ from the axis of the fiber in order to ensure total internal reflection within the fiber. Additional Reading: • “Optical fiber” on Wikipedia. • “Numerical aperture” on Wikipedia.

141

8.3 Dispersion in Optical Fiber [m0193]

Light may follow a variety of paths through a fiber optic cable. Each of the paths has a different length, leading to a phenomenon known as dispersion. Dispersion distorts signals and limits the data rate of digital signals sent over fiber optic cable. In this section, we analyze this dispersion and its effect on digital signals. Figure 8.5 shows the variety of paths that light may take through a straight fiber optic cable. The nominal path is shown in Figure 8.5(a), which is parallel to the axis of the cable. This path has the shortest associated propagation time. The path with the longest associated propagation time is shown in Figure 8.5(c). In this case, light bounces within the fiber, each time approaching the core-clad interface at the critical angle for total internal reflection. Any ray approaching at a greater angle is not completely reflected, and so would likely not survive to the end of

c S. Lally CC BY-SA 4.0

c S. Lally CC BY-SA 4.0

Figure 8.4: Cone of acceptance.

Figure 8.5: Paths that light may take through a straight fiber optic cable: (a) Along axis, corresponding to minimum path length; (b) Intermediate between (a) and (c); (c) Maximum length, corresponding to threshold angle for total internal reflection.

CHAPTER 8. OPTICAL FIBER

142

c S. Lally CC BY-SA 4.0

Figure 8.6: A digital signal that might be applied to the input of a fiber optic cable.

the cable. Figure 8.5(b) represents the continuum of possibilities between the extreme cases of (a) and (c), with associated propagation times greater than that of case (a) but less than that of case (c). Regardless of how light is inserted into the fiber, all possible paths depicted in Figure 8.5 are likely to exist. This is because fiber is rarely installed in a straight line, but rather follows a multiply-curved path. Each curve results in new angles of incidence upon the core-cladding boundary. The existence of these paths leads to dispersion. To see this, consider the input signal shown in Figure 8.6. This signal has period T , during which time a pulse of length ton < T may be present. Let the minimum propagation time through the fiber (as in Figure 8.5(a)) be τmin . Let the maximum propagation time through the fiber (as in Figure 8.5(c)) be τmax . If τmax − τmin ≪ ton , we see little degradation in the signal output from the fiber. Otherwise, we observe a “smearing” of pulses at the output of the fiber, as shown in Figure 8.7. Any smearing may pose problems for whatever device is detecting pulses at the receiving end. However, as τmax − τmin becomes a larger fraction of ton , we see that it becomes possible for adjacent pulses to overlap. This presents a much more serious challenge in detecting individual pulses, so let us examine the

c S. Lally CC BY-SA 4.0

Figure 8.7: The effect of dispersion on the signal output from the fiber optic cable: Top: Arriving waveform corresponding to minimum path length, Middle: Arriving waveform corresponding to maximum path length, and Bottom: Sum of all arriving waveforms.

8.3. DISPERSION IN OPTICAL FIBER

143

overlap problem in greater detail. To avoid overlap: τmax − τmin + ton < T

(8.18)

Let us define the quantity τ , τmax − τmin . This is sometimes referred to as the delay spread.1 Thus, we obtain the following requirement for overlap-free transmission: τ < T − ton (8.19) Note that the delay spread imposes a minimum value on T , which in turn imposes a maximum value on the rate at which information can be transmitted on the cable. So, we are motivated to calculate delay spread. The minimum propagation time tmin is simply the length l of the cable divided by the phase velocity vp of the light within the core; i.e., tmin = l/vp . Recall phase velocity is simply the speed of light in free √ space c divided by ǫr where ǫr is the relative permittivity of the core material. Thus: tmin =

l l l lnf = √ = = vp c/ ǫr c/nf c

(8.20)

The maximum propagation time tmax is different because the maximum path length is different. Specifically, the maximum path length is not l, but rather l/ cos θ2 where θ2 is the angle between the axis and the direction of travel as light crosses the axis. So, for example, θ2 = 0 for tmin since light in that case travels along the axis. The value of θ2 for tmax is determined by the threshold angle for total internal reflection. This was determined in Section 8.2 to be given by: cos θ2 =

nc nf

(threshold value)

so we obtain the following relationship: tmax =

ln2f l/ cos θ2 lnf /nc = = vp c/nf cnc

and subsequently we find: nf τ =l c 1 Full



 nf −1 nc

Note that τ increases linearly with l. Therefore, the rate at which pulses can be sent without overlap decreases linearly with increasing length. In practical applications, this means that the maximum supportable data rate decreases as the length of the cable increases. This is true independently of media loss within the cable (which we have not yet even considered!). Rather, this is a fundamental limitation resulting from dispersion. Example 8.3. Maximum supportable data rate in multimode fiber optic cable. A multimode fiber optic cable of length 1 m is used to transmit data using the scheme shown in Figure 8.6, with ton = T /2. Values of nf and nc for this fiber are 1.52 and 1.49, respectively. What is the maximum supportable data rate? Solution. Using Equation 8.23, we find the delay spread τ ∼ = 102 ps. To avoid overlap, T > 2τ ; therefore, T greater than about 204 ps is required. The modulation scheme allows one bit per period, so the maximum data rate is 1/T ∼ = 4.9 × 109 bits per second; i.e., ∼ = 4.9 Gb/s.

The finding of 4.9 Gb/s may seem like a pretty high data rate; however, consider what happens if the length increases to 1 km. It is apparent from Equation 8.23 that the delay spread will increase by a factor of 1000, so the maximum supportable data rate decreases by a factor of 1000 to a scant 4.9 Mb/s. Again, this is independent of any media loss within (8.21) the fiber. To restore the higher data rate over this longer path, the dispersion must be reduced. One way to do this is to divide the link into smaller links separated by repeaters which can receive the dispersed signal, demodulate it, regenerate the (8.22) original signal, and transmit the restored signal to the next repeater. Alternatively, one may employ “single mode” fiber, which has intrinsically less dispersion than the multimode fiber presumed in our analysis. (8.23)

disclosure: There are many ways to define “delay spread,” but this definition is not uncommon and is useful in the present analysis.

Additional Reading: • “Optical fiber” on Wikipedia. [m0209]

144

CHAPTER 8. OPTICAL FIBER

Image Credits c S. Lally, https://commons.wikimedia.org/wiki/File:Figure 8.1-01.svg, Fig. 8.1: CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). Modified by author. c Offaperry (S. Lally), Fig. 8.2: https://commons.wikimedia.org/wiki/File:Internal Reflection in Optical Fiber.svg, CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), Fig. 8.3: https://commons.wikimedia.org/wiki/File:Injecting Light into Optical Fiber.svg, CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), https://commons.wikimedia.org/wiki/File:Cone of Acceptance.svg, Fig. 8.4: CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), Fig. 8.5: https://commons.wikimedia.org/wiki/File:Paths of Light through Optic Cable.svg, CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), Fig. 8.6: https://commons.wikimedia.org/wiki/File:Digital Signal on Fiber Optic Cable.svg, CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), Fig. 8.7: https://commons.wikimedia.org/wiki/File:Dispersion on Signal Output from Fiber Optic Cable.svg, CC BY SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/).

Chapter 9

Radiation 9.1 Radiation from a Current Moment

units of A·m); and δ(r) is the volumetric sampling function1 defined as follows:

[m0194]

δ(r) , 0 for r 6= 0; and Z δ(r) dv , 1

(9.2) (9.3)

V

In this section, we begin to address the following problem: Given a distribution of impressed current density J(r), what is the resulting electric field intensity E(r)? One route to an answer is via Maxwell’s equations. Viewing Maxwell’s equations as a system of differential equations, a rigorous mathematical solution is possible given the appropriate boundary conditions. The rigorous solution following that approach is relatively complicated, and is presented beginning in Section 9.2 of this book. If we instead limit scope to a sufficiently simple current distribution, a simple informal derivation is possible. This section presents such a derivation. The advantage of tackling a simple special case first is that it will allow us to quickly assess the nature of the solution, which will turn out to be useful once we do eventually address the more general problem. Furthermore, the results presented in this section will turn out to be sufficient to tackle many commonly-encountered applications.

where V is any volume which includes the origin (r = 0). It is evident from Equation 9.3 that δ(r) has SI base units of m−3 . Subsequently, ∆J(r) has SI base units of A/m2 , confirming that it is a volume current density. However, it is the simplest possible form of volume current density, since – as indicated by Equation 9.2 – it exists only at the origin and nowhere else. Although some current distributions approximate the current moment, current distributions encountered in common engineering practice generally do not exist 1 Also a form of the Dirac delta function; see “Additional Reading” at the end of this section.

The simple current distribution considered in this section is known as a current moment. An example of a current moment is shown in Figure 9.1 and in this case is defined as follows: ˆ I ∆l δ(r) ∆J(r) = z

(9.1)

where I ∆l is the scalar component of the current moment, having units of current times length (SI base

c C. Wang CC BY-SA 4.0

Figure 9.1: A +ˆ z-directed current moment located at the origin.

c 2020 S.W. Ellingson CC BY SA 4.0. https://doi.org/10.21061/electromagnetics-vol-2 Electromagnetics Vol. 2.

CHAPTER 9. RADIATION

146 in precisely this form. Nevertheless, the current moment turns out to be generally useful as a “building block” from which practical distributions of current can be constructed, via the principle of superposition. Radiation from current distributions constructed in this manner is calculated simply by summing the radiation from each of the constituent current moments. Now let us consider the electric field intensity ∆E(r) that is created by this current distribution. First, if the current is steady (i.e., “DC”), this problem falls within the domain of magnetostatics; i.e., the outcome is completely described by the magnetic field, and there can be no radiation. Therefore, let us limit our attention to the “AC” case, for which radiation is possible. It will be convenient to employ phasor representation. In phasor representation, the current density is e ˆ Ie ∆l δ(r) ∆J(r) =z

(9.4)

where Ie ∆l is simply the scalar current moment expressed as a phasor. Now we are ready to address the question “What is e e ∆E(r) due to ∆J(r)?” Without doing any math, we e know quite a bit about ∆E(r). For example:

• Since electric fields are proportional to the e currents that give rise to them, we expect ∆E(r) e to be proportional to I ∆l .

• If we are sufficiently far from the origin, we e expect ∆E(r) to be approximately proportional to 1/r where r , |r| is the distance from the source current. This is because point sources give rise to spherical waves, and the power density in a spherical wave would be proportional to 1/r2 . Since time-average power e 2 e density is proportional to ∆E(r) , ∆E(r) must be proportional to 1/r.

• If we are sufficiently far from the origin, and the loss due to the medium is negligible, then we e expect the phase of ∆E(r) to change approximately at rate β where β is the phase propagation constant 2π/λ. Since we expect e spherical phasefronts, ∆E(r) should therefore −jβr contain the factor e .

ˆ-directed current • Ampere’s law indicates that a z ˆ at the origin should give rise to a φ-directed 2 magnetic field in the z = 0 plane. At the same time, Poynting’s theorem requires the cross product of the electric and magnetic fields to point in the direction of power flow. In the present problem, this direction is away from the e = 0) points in source; i.e., +ˆr. Therefore, ∆E(z the −ˆ z direction. The same principle applies outside of the z = 0 plane, so in general we e expect ∆E(r) to point in the θˆ direction. e • We expect ∆E(r) = 0 along the z axis. e Subsequently ∆E(ˆ r) must increase from zero at θ = 0 and return to zero at θ = π. The e symmetry of the problem suggests ∆E(ˆ r) is

maximum at θ = π/2. This magnitude must vary in the simplest possible way, leading us to e r) is proportional to sin θ. conclude that ∆E(ˆ Furthermore, the radial symmetry of the problem e r) should not depend at all on φ. means that ∆E(ˆ

Putting these ideas together, we conclude that the radiated electric field has the following form:   −jβr ˆ Ie ∆l (sin θ) e e ∆E(r) ≈ θC r

(9.5)

where C is a constant which accounts for all of the constants of proportionality identified in the preceding e analysis. Since the units of ∆E(r) are V/m, the units of C must be Ω/m. We have not yet accounted for the wave impedance of the medium η, which has units of Ω, so it would be a good bet based on the units that C is proportional to η. However, here the informal analysis reaches a dead end, so we shall simply state the result from the rigorous solution: C = jηβ/4π. The units are correct, and we finally obtain: jηβ  e  e−jβr e ∆E(r) ≈ θˆ I ∆l (sin θ) 4π r

(9.6)

Additional evidence that this solution is correct comes from the fact that it satisfies the wave equation e e ∇2 ∆E(r) + β 2 ∆E(r) = 0.3

2 This is sometimes described as the “right hand rule” of Ampere’s law. 3 Confirming this is straightforward (simply substitute and evaluate) and is left as an exercise for the student.

9.2. MAGNETIC VECTOR POTENTIAL Note that the expression we have obtained for the radiated electric field is approximate (hence the “≈”). This is due in part to our presumption of a simple spherical wave, which may only be valid at distances far from the source. But how far? An educated guess would be distances much greater than a wavelength (i.e., r ≫ λ). This will do for now; in another section, we shall show rigorously that this guess is essentially correct. We conclude this section by noting that the current distribution analyzed in this section is sometimes referred to as a Hertzian dipole. A Hertzian dipole is typically defined as a straight infinitesimally-thin filament of current with length which is very small relative to a wavelength, but not precisely zero. This interpretation does not change the solution obtained in this section, thus we may view the current moment and the Hertzian dipole as effectively the same in practical engineering applications. Additional Reading: • “Dirac delta function” on Wikipedia. • “Dipole antenna” (section entitled “Hertzian Dipole”) on Wikipedia.

147

9.2 Magnetic Vector Potential [m0195]

A common problem in electromagnetics is to determine the fields radiated by a specified current distribution. This problem can be solved using Maxwell’s equations along with the appropriate electromagnetic boundary conditions. For time-harmonic (sinusoidally-varying) currents, we use phasor representation.4 Given the specified e and the desired electromagnetic current distribution J e and H, e the appropriate equations are: fields E e = ρev /ǫ ∇·E

(9.7)

e =0 ∇·H

(9.9)

e = −jωµH e ∇×E

(9.8)

e =J e + jωǫE e ∇×H

(9.10)

e =0 ∇·E

(9.11)

where ρev is the volume charge density. In most engineering problems, one is concerned with propagation through media which are well-modeled as homogeneous media with neutral charge, such as free space.5 Therefore, in this section, we shall limit our scope to problems in which ρev = 0. Thus, Equation 9.7 simplifies to:

To solve the linear system of partial differential Equations 9.8–9.11, it is useful to invoke the concept of magnetic vector potential. The magnetic vector potential is a vector field that has the useful property that it is able to represent both the electric and magnetic fields as a single field. This allows the formidable system of equations identified above to be reduced to a single equation which is simpler to solve. Furthermore, this single equation turns out to be the wave equation, with the slight difference that the equation will be mathematically inhomogeneous, with the inhomogeneous part representing the source current. 4 Recall that there is no loss of generality in doing so, since any other time-domain variation in the current distribution can be represented using sums of time-harmonic solutions via the Fourier transform. 5 A counter-example would be propagation through a plasma, which by definition consists of non-zero net charge.

CHAPTER 9. RADIATION

148 e is defined by the The magnetic vector potential A following relationship:

Astute readers might already realize what we’re up to here. Equation 9.16 is very similar to the relationship E = −∇V from electrostatics,6 in which V is the e e B,∇×A (9.12) scalar electric potential field. Evidently Equation 9.16 is an enhanced version of that relationship that e = µH e is the magnetic flux density. The accounts for the coupling with H (here, represented where B by A) in the time-varying (decidedly non-static) case. magnetic field appears in three of Maxwell’s That assessment is correct, but let’s not get too far equations. For Equation 9.12 to be a reasonable e must yield reasonable results when ahead of ourselves: As demonstrated in the previous definition, ∇ × A e in each of these equations. Let us paragraph, we are not yet compelled to make any substituted for µH particular choice for Ve , and this freedom will be first check for consistency with Gauss’ law for exploited later in this section. magnetic fields, Equation 9.9. Making the substitution, we obtain: Next we check for consistency with Equation 9.10.   e ∇· ∇×A =0 (9.13) Making the substitution:   1 e e + jωǫE e ∇×A =J (9.19) ∇× This turns out to be a mathematical identity that µ applies to any vector field (see Equation B.24 in Appendix B.3). Therefore, Equation 9.12 is Multiplying both sides of the equation by µ: consistent with Gauss’ law for magnetic fields. e = µJ e + jωµǫE e ∇×∇×A (9.20) Next we check for consistency with Equation 9.8. e yielding: Next we use Equation 9.16 to eliminate E, Making the substitution:     e = µJ e + jωµǫ −∇Ve − jω A e (9.21) ∇×∇×A e e ∇ × E = −jω ∇ × A (9.14) Gathering terms on the left, we obtain   e + jω A e =0 ∇× E

After a bit of algebra, we obtain

(9.15)

Now, for reasons that will become apparent in just a moment, we define a new scalar field Ve and require it to satisfy the following relationship: e + jω A e − ∇Ve , E

e = ω 2 µǫA e − jωµǫ∇Ve + µJ e ∇×∇×A

(9.22)

Now we replace the left side of this equation using vector identity B.29 in Appendix B.3:   e ≡∇ ∇·A e − ∇2 A e ∇×∇×A (9.23)

(9.16) Equation 9.22 becomes:   e − ∇2 A e = ω 2 µǫA e − jωµǫ∇Ve + µJ e ∇ ∇·A Using this definition, Equation 9.15 becomes:   (9.24) ∇ × −∇Ve = 0 (9.17) Now multiplying both sides by −1 and rearranging terms:   which is simply e + ω 2 µǫA e =∇ ∇·A e + jωµǫ∇Ve − µJ e ∇2 A ∇ × ∇Ve = 0 (9.18) (9.25) Combining terms on the right side: Once again we have obtained a mathematical identity   that applies to any vector field (see Equation B.25 in 2e 2 e =∇ ∇·A e + jωµǫVe −µJ e (9.26) ∇ A+ω µǫ A Appendix B.3). Therefore, Ve can be any mathematically-valid scalar field. Subsequently, e + jωµǫVe Now consider the expression ∇ · A Equation 9.12 is consistent with Equation 9.8 appearing in the parentheses on the right side of the (Maxwell’s curl equation for the electric field) for any 6 Note: No tilde in this expression. choice of Ve that we are inclined to make.

9.2. MAGNETIC VECTOR POTENTIAL

149

equation. We established earlier that Ve can be essentially any scalar field – from a mathematical perspective, we are free to choose. Invoking this freedom, we now require Ve to satisfy the following expression: e + jωµǫVe = 0 ∇·A (9.27) Clearly this is advantageous in the sense that Equation 9.26 is now dramatically simplified. This equation becomes: e + ω 2 µǫA e = −µJ e ∇2 A

(9.28)

Note that this expression is a wave equation. In fact e and H e it is the same wave equation that determines E in source-free regions, except the right-hand side is not zero. Using mathematical terminology, we have e in the form of an obtained an equation for A inhomogeneous partial differential equation, where the inhomogeneous part includes – no surprise here – e the source current J. Now we have what we need to find the electromagnetic fields radiated by a current distribution. The procedure is simply as follows:

e 1. Solve the partial differential Equation 9.28 for A along with the appropriate electromagnetic boundary conditions. e = (1/µ)∇ × A e 2. H

e may now be determined from H e using 3. E Equation 9.10. Summarizing: e is a vector field, The magnetic vector potential A defined by Equation 9.12, that is able to represent both the electric and magnetic fields simultaneously.

Also: To determine the electromagnetic fields radiated e one may solve Equaby a current distribution J, e tion 9.28 for A and then use Equation 9.12 to e and subsequently E. e determine H

Specific techniques for performing this procedure – in particular, for solving the differential equation – vary depending on the problem, and are discussed in other sections of this book. We conclude this section with a few comments about Equation 9.27. This equation is known as the Lorenz gauge condition. This constraint is not quite as arbitrary as the preceding derivation implies; rather, there is some deep physics at work here. Specifically, the Lorenz gauge leads to the classical interpretation of Ve as the familiar scalar electric potential, as noted previously in this section. (For additional information on that idea, recommended starting points are included in “Additional Reading” at the end of this section.) At this point, it should be clear that the electric and magnetic fields are not merely coupled quantities, but in fact two aspects of the same field; namely, the magnetic vector potential. In fact modern physics (quantum mechanics) yields the magnetic vector potential as a description of the “electromagnetic force,” a single entity which constitutes one of the four fundamental forces recognized in modern physics; the others being gravity, the strong nuclear force, and the weak nuclear force. For more information on that concept, an excellent starting point is the video “Quantum Invariance & The Origin of The Standard Model” referenced at the end of this section. Additional Reading: • “Lorenz gauge condition” on Wikipedia. • “Magnetic potential” on Wikipedia. • PBS Space Time video “Quantum Invariance & The Origin of The Standard Model,” available on YouTube.

CHAPTER 9. RADIATION

150

9.3 Solution of the Wave Equation for Magnetic Vector Potential

direction of current flow, and δ(r) is the volumetric sampling function8 defined as follows: δ(r) , 0 for r 6= 0; and Z δ(r) dv , 1

(9.32) (9.33)

V

[m0196]

where V is any volume which includes the origin (r = 0). It is evident from Equation 9.33 that δ(r) has SI base units of m−3 ; i.e., inverse volume. 2 e − γ2A e = −µJ e ∇2 A (9.29) Subsequently J(r) has SI base units of A/m , indicating that it is a volume current density.9 – as where γ is the propagation constant, defined in the indicated by Equation 9.32 – it exists only at the usual manner7 origin and nowhere else. γ 2 , −ω 2 µǫ (9.30) Substituting Equation 9.31 into Equation 9.29, we Equation 9.29 is a partial differential equation which obtain: is inhomogeneous (in the mathematical sense) and e due to a current The magnetic vector potential A e is given by the following wave equation: density J

can be solved given appropriate boundary conditions. In this section, we present the solution for arbitrary distributions of current in free space. The strategy is to first identify a solution for a distribution of current that exists at a single point. This distribution is the current moment. An example of a current moment is shown in Figure 9.2. A general expression for a current moment located at the origin is: e J(r) = ˆl Ie ∆l δ(r)

e − γ2A e = −µˆl Ie ∆l δ(r) ∇2 A

(9.34)

The general solution to this equation presuming homogeneous and time-invariant media (i.e., µ and ǫ constant with respect to space and time) is: e±γr e A(r) = ˆl µ Ie ∆l 4πr

(9.35)

To confirm that this is a solution to Equation 9.34, substitute this expression into Equation 9.29 and (9.31) observe that the equality holds.

where Ie has units of current (SI base units of A), ∆l has units of length (SI base units of m), ˆl is the

7 Alternatively, γ 2 , −ω 2 µǫ accounting for the possibility of c lossy media.

Note that Equation 9.35 indicates two solutions, corresponding to the signs of the exponent in the factor e±γr . We can actually be a little more specific by applying some common-sense physics. Recall that γ in general may be expressed in terms of real and imaginary components as follows: γ = α + jβ

(9.36)

where α is a real-valued positive constant known as the attenuation constant and β is a real-valued positive constant known as the phase propagation constant. Thus, we may rewrite Equation 9.35 as follows:

c C. Wang CC BY-SA 4.0

Figure 9.2: An example of a current moment located ˆ. at the origin. In this case, ˆl = z

e±αr e±jβr e A(r) = ˆl µ Ie ∆l 4πr

(9.37)

8 Also a form of the Dirac delta function; see “Additional Reading” at the end of this section. 9 Remember, the density of a volume current is with respect to the area through which it flows, therefore the units are A/m2 .

9.3. SOLUTION OF THE WAVE EQUATION FOR MAGNETIC VECTOR POTENTIAL Now consider the factor e±αr /r, which determines the dependence of magnitude on distance r. If we choose the negative sign in the exponent, this factor decays exponentially with increasing distance from the origin, ultimately reaching zero at r → ∞. This is precisely the expected behavior, since we expect the magnitude of a radiated field to diminish with increasing distance from the source. If on the other hand we choose the positive sign in the exponent, this factor increases to infinity as r → ∞. That outcome can be ruled out on physical grounds, since it implies the introduction of energy independently from the source. The requirement that field magnitude diminishes to zero as distance from the source increases to infinity is known as the radiation condition, and is essentially a boundary condition that applies at r → ∞. Invoking the radiation condition, Equation 9.35 becomes: e−γr e A(r) = ˆl µ Ie ∆l 4πr

(9.38)

In the loss-free (α = 0) case, we cannot rely on the radiation condition to constrain the sign of γ. However, in practical engineering work there is always some media loss; i.e., α might be negligible but is not quite zero. Thus, we normally assume that the solution for lossless (including free space) conditions is given by Equation 9.38 with α = 0: −jβr

e e A(r) = ˆl µ Ie ∆l 4πr

(9.39)

Now consider the factor e−jβr in Equation 9.39. This factor exclusively determines the dependence of the e phase of A(r) with increasing distance r from the source. In this case, we observe that surfaces of constant phase correspond to spherical shells which e are concentric with the source. Thus, A(r) is a spherical wave. Let us now consider a slightly more complicated version of the problem in which the current moment is no longer located at the origin, but rather is located at r′ . This is illustrated in Figure 9.3. This current distribution can be expressed as follows: e J(r) = ˆl Ie ∆l δ(r − r′ )

(9.40)

The solution in this case amounts to a straightforward modification of the existing solution. To see this, note

151

c C. Wang CC BY-SA 4.0

Figure 9.3: Current moment displaced from the origin.

e depends only on r, and not at all on θ or φ. In that A e depends only on the distance between other words, A e and the the “field point” r at which we observe A, ′ “source point” r at which the current moment lies. Thus we replace r with this distance, which is |r − r′ |. The solution becomes: e−γ |r−r | e A(r) = ˆl µ Ie ∆l 4π |r − r′ | ′

(9.41)

Continuing to generalize the solution, let us now consider a scenario consisting of a filament of current following a path C through space. This is illustrated in Figure 9.4. Such a filament may be viewed as a collection of a large number N of discrete current moments distributed along the path. The contribution of the nth current moment, located at rn , to the total magnetic vector potential is:

e rn ) = ˆl(rn ) µ I(r e n ) ∆l ∆A(r;

e−γ|r−rn | (9.42) 4π |r − rn |

Note that we are allowing both the current Ie and current direction ˆl to vary with position along C. Assuming the medium is linear, superposition applies. So we add these contributions to obtain the total

CHAPTER 9. RADIATION

152

9.4 Radiation from a Hertzian Dipole [m0197]

c C. Wang CC BY-SA 4.0

Section 9.1 presented an informal derivation of the electromagnetic field radiated by a Hertzian dipole represented by a zero-length current moment. In this section, we provide a rigorous derivation using the concept of magnetic vector potential discussed in Sections 9.2 and 9.3. A review of those sections is recommended before tackling this section.

Figure 9.4: A filament of current lying along the path A Hertzian dipole is commonly defined as an C. This current distribution may be interpreted as a electrically-short and infinitesimally-thin straight collection of current moments lying along C. filament of current, in which the density of the current is uniform over its length. The Hertzian dipole is commonly used as a “building block” for constructing magnetic vector potential: physically-realizable distributions of current as N X exhibited by devices such as wire antennas. The e e rn ) A(r) ≈ ∆A(r; (9.43) method is to model these relatively complex n=1 distributions of current as the sum of Hertzian dipoles, N −γ|r−rn | µ Xˆ e which reduces the problem to that of summing the e n) ≈ l(rn ) I(r ∆l (9.44) contributions of the individual Hertzian dipoles, with 4π n=1 |r − rn | each Hertzian dipole having the appropriate (i.e., Now letting ∆l → 0 so that we may replace ∆l with different) position, magnitude, and phase. the differential length dl: To facilitate use of the Hertzian dipole as a building Z −γ |r−r′ | block suitable for constructing physically-realizable e µ ˆl(r′ ) I(r e e ′) A(r) = dl (9.45) distributions of current, we choose to represent the 4π C |r − r′ | Hertzian dipole using an essentially equivalent current distribution which is mathematically more versatile. where we have also replaced rn with the original ′ notation r since we once again have a continuum (as This description of the Hertzian dipole replaces the notion of constant current over finite length with the opposed to a discrete set) of source locations. notion of a current moment located at a single point. The magnetic vector potential corresponding to This is shown in Figure 9.5, and is given by: radiation from a line distribution of current is e ∆J(r) = ˆl Ie ∆l δ(r) (9.46) given by Equation 9.45. e (SI base units of A·m) is the where the product I∆l ˆl is the direction of current flow, and current moment, e Given A(r), the magnetic and electric fields may be δ(r) is the volumetric sampling function defined as determined using the procedure developed in follows: Section 9.2. δ(r) , 0 for r 6= 0; and (9.47) Z Additional Reading: δ(r) dv , 1 (9.48) V

• “Magnetic potential” on Wikipedia.

• “Dirac delta function” on Wikipedia.

where V is any volume which includes the origin (r = 0). In this description, the Hertzian dipole is located at the origin.

9.4. RADIATION FROM A HERTZIAN DIPOLE

153

The solution for the magnetic vector potential due to a At this point, it is convenient to make the following ˆ-directed Hertzian dipole located at the origin was z definitions: presented in Section 9.3. In the present scenario, it is: e−jβr C , (cos θ) (9.56) r r e−γr e ˆ µ Ie ∆l (9.49) A(r) =z e−jβr 4πr (9.57) Cθ , − (sin θ) r where the propagation constant γ = α + jβ as usual. These definitions allow Equation 9.55 to be written Assuming lossless media (α = 0), we have compactly as follows: e−jβr e ˆ µ Ie ∆l A(r) =z (9.50) h i e−jβr 4πr ˆ θ ˆ = ∇ × ˆrCr + θC (9.58) ∇×z r We obtain the magnetic field intensity using the The right side of Equation 9.58 is evaluated using definition of magnetic vector potential: Equation B.18 (Appendix B.2). Although the e , (1/µ)∇ × A e H (9.51) complete expression consists of 6 terms, only 2 terms are non-zero.10 This leaves: Ie ∆l e−jβr ˆ = ∇×z (9.52)   4π r e−jβr 1 ∂ ∂ ˆ ˆ ∇×z =φ (rCθ ) − Cr (9.59) r r ∂r ∂θ ˆ into the spherical To proceed, it is useful to convert z   −jβr coordinate system. To do this, we find the component 1 ˆ (sin θ) e = φ jβ + (9.60) ˆ ˆ ˆ that is parallel to ˆr, θ, and φ; and then sum the of z r r results: Substituting this result into Equation 9.52, we obtain:     ˆ = ˆr (ˆr · z ˆ) + θˆ θˆ · z ˆ + φˆ φˆ · z ˆ z (9.53)   −jβr e 1 e = φˆ I ∆l (sin θ) e jβ + (9.61) H ˆ = ˆr cos θ − θ sin θ + 0 (9.54) 4π r r Equation 9.52 requires computation of the following quantity:  e−jβr e−jβr ˆ ∇×z = ∇ × ˆr (cos θ) r r  −jβr e ˆ −θ (sin θ) (9.55) r

Let us further limit our scope to the field far from the antenna. Specifically, let us assume r ≫ λ. Now we use the relationship β = 2π/λ and determine the following: jβ +

1 2π 1 =j + r λ r 2π = jβ ≈j λ

(9.62) (9.63)

Equation 9.61 becomes: −jβr e ˆ I · β∆l (sin θ) e e ≈ φj H 4π r

(9.64)

where the approximation holds for low-loss media and r ≫ λ. This expression is known as a far field approximation, since it is valid only for distances “far” (relative to a wavelength) from the source. Now let us take a moment to interpret this result: c C. Wang CC BY-SA 4.0

Figure 9.5: A Hertzian dipole located at the origin, ˆ. represented as a current moment. In this case, ˆl = z

10 Specifically, two terms are zero because there is no φ ˆ component in the argument of the curl function; and another two terms are zero because the argument of the curl function is independent of φ, so partial derivatives with respect to φ are zero.

CHAPTER 9. RADIATION

154 • Notice the factor β∆l has units of radians; that is, it is electrical length. This tells us that the magnitude of the radiated field depends on the electrical length of the current moment. • The factor e−jβr /r indicates that this is a spherical wave; that is, surfaces of constant phase correspond to concentric spheres centered on the source, and magnitude is inversely proportional to distance. • The direction of the magnetic field vector is ˆ which is precisely what we expect; for always φ, example, using the Biot-Savart law. • Finally, note the factor sin θ. This indicates that the field magnitude is zero along the direction in which the source current flows, and is a maximum in the plane perpendicular to this direction. Now let us determine the electric field radiated by the Hertzian dipole. The direct method is to employ Ampere’s law. That is, e e = 1 ∇×H E jωǫ

(9.65)

e is given by Equation 9.64. At field points far where H from the dipole, the radius of curvature of the spherical phasefronts is very large and so appear to be locally planar. That is, from the perspective of an observer far from the dipole, the arriving wave appears to be a plane wave. In this case, we may employ the plane wave relationships. The appropriate relationship in this case is: e = −ηˆr × H e E

(9.66)

where η is the wave impedance. So we find: −jβr e ˆ I · β∆l (sin θ) e e ≈ θjη E 4π r

(9.67)

Summarizing:

The electric and magnetic fields far (i.e., ≫ λ) ˆ-directed Hertzian dipole having confrom a z stant current Ie over length ∆l, located at the origin, are given by Equations 9.67 and 9.64, respectively.

Additional Reading: • “Dipole antenna” (section entitled “Hertzian Dipole”) on Wikipedia.

9.5. RADIATION FROM AN ELECTRICALLY-SHORT DIPOLE

155

9.5 Radiation from an Electrically-Short Dipole [m0198]

The simplest distribution of radiating current that is encountered in common practice is the electrically-short dipole (ESD). This current distribution is shown in Figure 9.6. The two characteristics that define the ESD are (1) the current is aligned along a straight line, and (2) the length L of the line is much less than one-half of a wavelength; i.e., L ≪ λ/2. The latter characteristic is what we mean by “electrically-short.”11 The current distribution of an ESD is approximately triangular in magnitude, and approximately constant in phase. How do we know this? First, note that distributions of current cannot change in a complex or rapid way over such distances which are much less than a wavelength. If this is not immediately apparent, recall the behavior of transmission lines: 11 A potential source of confusion is that the Hertzian dipole is also a “dipole” which is “electrically-short.” The distinction is that the current comprising a Hertzian dipole is constant over its length. This condition is rarely and only approximately seen in practice, whereas the triangular magnitude distribution is a relatively good approximation to a broad class of commonly-encountered electricallyshort wire antennas. Thus, the term “electrically-short dipole,” as used in this book, refers to the triangular distribution unless noted otherwise.

c C. Wang CC BY-SA 4.0

Figure 9.7: Current distribution of the electricallyshort dipole (ESD) approximated as a large number of Hertzian dipoles. The current standing wave on a transmission line exhibits a period of λ/2, regardless the source or termination. For the ESD, L ≪ λ/2 and so we expect an even simpler variation. Also, we know that the current at the ends of the dipole must be zero, simply because the dipole ends there. These considerations imply that the current distribution of the ESD is well-approximated as triangular in magnitude.12 Expressed mathematically:   2 e (9.68) I(z) ≈ I0 1 − |z| L

where I0 (SI base units of A) is a complex-valued constant indicating the maximum current magnitude and phase.

c C. Wang CC BY-SA 4.0

Figure 9.6: Current distribution of the electricallyshort dipole (ESD).

There are two approaches that we might consider in order to find the electric field radiated by an ESD. The first approach is to calculate the magnetic vector e by integration over the current potential A e = (1/µ)∇ × A, e and finally distribution, calculate H e e calculate E from H using Ampere’s law. We shall employ a simpler approach, shown in Figure 9.7. Imagine the ESD as a collection of many shorter segments of current that radiate independently. The total field is then the sum of these short segments. Because these segments are very short relative to the 12 A more rigorous analysis leading to the same conclusion is possible, but is beyond the scope of this book.

CHAPTER 9. RADIATION

156 length of the dipole as well as being short relative to a wavelength, we may approximate the current over each segment as approximately constant. In other words, we may interpret each of these segments as being, to a good approximation, a Hertzian dipole. The advantage of this approach is that we already have a solution for each of the segments. In ˆ-directed Hertzian Section 9.4, it is shown that a z dipole at the origin radiates the electric field −jβr e ˆ I · β∆l (sin θ) e e E(r) ≈ θjη 4π r

(9.69) c C. Wang CC BY-SA 4.0

where Ie and ∆l may be interpreted as the current and length of the dipole, respectively. In this expression, η Figure 9.8: Parallel ray approximation for an ESD. is the wave impedance of medium in which the dipole radiates (e.g., ≈ 377 Ω for free space), and we have yielding: presumed lossless media such that the attenuation Z −jβ |r−ˆ zz ′ | constant α ≈ 0 and the phase propagation constant ηβ +L/2 ˆ′ e ′ ′ e e θ I(z ) (sin θ ) dz ′ E(r) ≈ j β = 2π/λ. This expression also assumes field points ˆz ′ | 4π −L/2 |r − z far from the dipole; specifically, distances r that are (9.74) much greater than λ. Repurposing this expression for Given some of the assumptions we have already the present problem, the segment at the origin radiates made, this expression can be further simplified. For the electric field: example, note that θ′ ≈ θ since L ≪ r. For the same ˆ Since these variables are −jβr reason, θˆ′ ≈ θ. ˆ I0 · β∆l (sin θ) e e z ′ = 0) ≈ θjη (9.70) approximately constant over the length of the dipole, E(r; 4π r we may move them outside the integral, yielding: where the notation z ′ = 0 indicates the Hertzian ′ Z +L/2 dipole is located at the origin. Letting the length ∆l e−jβ |r−ˆzz | ′ ηβ ′ ˆ e e E(r) ≈ θj (sin θ) I(z ) dz of this segment shrink to differential length dz ′ , we ˆz ′ | 4π |r − z −L/2 may describe the contribution of this segment to the (9.75) field radiated by the ESD as follows: ˆz ′ |. It is also possible to simplify the expression |r − z ′ −jβr e I · βdz 0 ′ ˆ e z = 0) ≈ θjη (sin θ) (9.71) Consider Figure 9.8. Since we have already assumed dE(r; 4π r that r ≫ L (i.e., the distance to field points is much Using this approach, the electric field radiated by any greater than the length of the dipole), the vector r is ˆz ′ . approximately parallel to the vector r − z segment can be written: Subsequently, it must be true that −jβ |r−ˆ zz ′ | e ′) I(z e e z ′ ) ≈ θˆ′ jηβ dE(r; (sin θ′ ) dz ′ ˆz ′ | ≈ r − ˆr · z ˆz ′ |r − z (9.76) ˆz ′ | 4π |r − z (9.72) Note that the magnitude of r − ˆr · z ˆz ′ must be ˆz ′ Note that θ is replaced by θ′ since the ray r − z approximately equal to r, since r ≫ L. So, insofar as ˆ. forms a different angle (i.e., θ′ ) with respect to z e ˆz ′ | determines the magnitude of E(r), |r − z we may Similarly, θˆ is replaced by θˆ′ , since it also varies with use the approximation: z ′ . The electric field radiated by the ESD is obtained by integration over these contributions: ˆz ′ | ≈ r (magnitude) |r − z (9.77) Z +L/2 ˆz ′ | determines phase, we have to be a Insofar as |r − z e e r; z ′ ) E(r) ≈ dE(ˆ (9.73) bit more careful. The part of the integrand of −L/2

9.5. RADIATION FROM AN ELECTRICALLY-SHORT DIPOLE Equation 9.75 that exhibits varying phase is ′ e−jβ |r−ˆzz | . Using Equation 9.76, we find e−jβ |r−ˆzz | ≈ e−jβr e+jβˆr·ˆzz ′



physically realizable, whereas Hertzian dipoles are not. (9.78)

The worst case in terms of phase variation within the integral is for field points along the z axis. For these ˆ = ±1 and subsequently |r − z ˆz ′ | varies points, ˆr · z from z − L/2 to z + L/2 where z is the location of the field point. However, since L ≪ λ (i.e., because the dipole is electrically short), this difference in lengths is much less than λ/2. Therefore, the phase ˆz ′ varies by much less than π radians, and βˆr · z ′ subsequently e−jβˆr·ˆzz ≈ 1. We conclude that under these conditions, e−jβ |r−ˆzz | ≈ e−jβr (phase) ′

(9.79)

Applying these simplifications for magnitude and phase to Equation 9.75, we obtain: −jβr

ˆ ηβ (sin θ) e e E(r) ≈ θj 4π r

Z

+L/2 −L/2

e ′ )dz ′ I(z

(9.80) The integral in this equation is very easy to evaluate; in fact, from inspection (Figure 9.6), we determine it is equal to I0 L/2. Finally, we obtain: −jβr ˆ I0 · βL (sin θ) e e E(r) ≈ θjη 8π r

157

(9.81)

Summarizing:

The electric field intensity radiated by an ESD located at the origin and aligned along the z axis is given by Equation 9.81. This expression is valid for r ≫ λ. It is worth noting that the variation in magnitude, phase, and polarization of the ESD with field point location is identical to that of a single Hertzian dipole ˆI0 L/2 (Section 9.4). having current moment z However, the magnitude of the field radiated by the ESD is exactly one-half that of the Hertzian dipole. Why one-half? Simply because the integral over the triangular current distribution assumed for the ESD is one-half the integral over the uniform current distribution that defines the Hertzian dipole. This similarly sometimes causes confusion between Hertzian dipoles and ESDs. Remember that ESDs are

It is common to eliminate the factor of β in the magnitude using the relationship β = 2π/λ, yielding: −jβr ˆ ηI0 L (sin θ) e e E(r) ≈ θj 4 λ r

(9.82)

At field points r ≫ λ, the wave appears to be locally planar. Therefore, we are justified using the plane e = 1 ˆr × E e to calculate H. e The wave relationship H η result is: −jβr ˆ I0 L (sin θ) e e H(r) ≈ φj 4 λ r

(9.83)

Finally, let us consider the spatial characteristics of the radiated field. Figures 9.9 and 9.10 show the result in a plane of constant φ. Figures 9.11 and 9.12 show the result in the z = 0 plane. Note that the orientations of the electric and magnetic field vectors e ×H e that is always indicate a Poynting vector E directed radially outward from the location of the dipole. This confirms that power flow is always directed radially outward from the dipole. Due to the symmetry of the problem, Figures 9.9–9.12 provide a complete characterization of the relative magnitudes and orientations of the radiated fields.

CHAPTER 9. RADIATION

158

c S. Lally CC BY-SA 4.0

c S. Lally CC BY-SA 4.0

Figure 9.9: Magnitude of the radiated field in any Figure 9.11: Magnitude of the radiated field in any plane of constant φ. plane of constant z.

c S. Lally CC BY-SA 4.0

c S. Lally CC BY-SA 4.0

Figure 9.10: Orientation of the electric and magnetic Figure 9.12: Orientation of the electric and magnetic fields in any plane of constant φ. fields in the z = 0 plane.

9.6. FAR-FIELD RADIATION FROM A THIN STRAIGHT FILAMENT OF CURRENT

159

9.6 Far-Field Radiation from a Thin Straight Filament of Current [m0199]

A simple distribution of radiating current that is encountered in common practice is the thin straight current filament, shown in Figure 9.13. The defining characteristic of this distribution is that the current filament is aligned along a straight line, and that the maximum dimension of the cross-section of the filament is much less than a wavelength. The latter characteristic is what we mean by “thin” – i.e., the filament is “electrically thin.” Physical distributions of current that meet this description include the electrically-short dipole (Section 9.5) and the half-wave dipole (Section 9.7) among others. A gentler introduction to this distribution is via Section 9.5 (“Radiation from an Electrically-Short Dipole”), so students may want to review that section first. This section presents the more general case. Let the magnitude and phase of current along the e (SI base filament be given by the phasor quantity I(z) e units of A). In principle, the only constraint on I(z)

c C. Wang CC BY-SA 4.0

c C. Wang CC BY-SA 4.0

Figure 9.14: Current distribution approximated as a set of Hertzian dipoles.

that applies generally is that it must be zero at the e = 0 for |z| ≥ L/2. ends of the distribution; i.e., I(z) Details beyond this constraint depend on L and perhaps other factors. Nevertheless, the characteristics of radiation from members of this class of current distributions have much in common, regardless of L. These become especially apparent if we limit our scope to field points far from the current, as we shall do here. There are two approaches that we might consider in order to find the electric field radiated by this distribution. The first approach is to calculate the e by integration over the magnetic vector potential A current distribution (Section 9.3), calculate e = (1/µ)∇ × A, e and finally calculate E e from H e H using the differential form of Ampere’s law. In this section, we shall employ a simpler approach, shown in Figure 9.14. Imagine the filament as a collection of many shorter segments of current that radiate independently. The total field is then the sum of these short segments. Because these segments are very short relative to the length of the filament as well as being short relative to a wavelength, we may approximate the current over each segment as constant. In other words, we may interpret each segment as being, to a good approximation, a Hertzian dipole.

Figure 9.13: A thin straight distribution of radiating The advantage of this approach is that we already current. have a solution for every segment. In Section 9.4, it is

CHAPTER 9. RADIATION

160 ˆ-directed Hertzian dipole at the origin shown that a z radiates the electric field −jβr e ˆ I (β∆l) (sin θ) e e E(r) ≈ θjη 4π r

(9.84)

where Ie and ∆l may be interpreted as the current and length of the filament, respectively. In this expression, η is the wave impedance of medium in which the filament radiates (e.g., ≈ 377 Ω for free space), and we have presumed lossless media such that the attenuation constant α ≈ 0 and the phase propagation constant β = 2π/λ. This expression also assumes field points far from the filament; specifically, distances r that are much greater than λ. Repurposing this expression for the present problem, we note that the segment at the origin radiates the electric field: −jβr e ˆ I(0) (β∆l) (sin θ) e e z ′ = 0) ≈ θjη E(r; 4π r (9.85) where the notation z ′ = 0 indicates the Hertzian dipole is located at the origin. Letting the length ∆l of this segment shrink to differential length dz ′ , we may describe the contribution of this segment to the field radiated by the ESD as follows:

c C. Wang CC BY-SA 4.0

Figure 9.15: Parallel ray approximation. Given some of the assumptions we have already made, this expression can be further simplified. For example, note that θ′ ≈ θ since L ≪ r. For the same ˆ Since these variables are reason, θˆ′ ≈ θ. approximately constant over the length of the filament, we may move them outside the integral, yielding: Z +L/2 −jβ |r−ˆ zz ′ | ηβ ′ e ˆ e e E(r) ≈ θj (sin θ) I(z ) dz ′ ˆz ′ | 4π |r − z −L/2 (9.90)

−jβr ′ e ˆ I(0) (βdz ) (sin θ) e e z ′ = 0) ≈ θjη dE(r; 4π r (9.86) ˆz ′ |. It is also possible to simplify the expression |r − z Using this approach, the electric field radiated by any Consider Figure 9.15. Since we have already assumed segment can be written: that r ≫ L (i.e., the distance to field points is much greater than the length of the filament), the vector r is −jβ |r−ˆ zz ′ | ′ e ′ e z ′ ) ≈ θˆ′ jηβ I(z ) (sin θ′ ) e ˆz ′ . approximately parallel to the vector r − z dE(r; dz ˆz ′ | 4π |r − z Subsequently, it must be true that (9.87) ′ ′ ˆz Note that θ is replaced by θ since the ray r − z ˆz ′ | ≈ r − ˆr · z ˆz ′ |r − z (9.91) ′ ′ ˆ. forms a different angle (i.e., θ ) with respect to z ≈ r − z cos θ (9.92) Subsequently, θˆ is replaced by θˆ′ , which varies ˆz ′ must be Note that the magnitude of r − ˆr · z similarly with z ′ . The electric field radiated by the approximately equal to r, since r ≫ L. So, insofar as filament is obtained by integration over these e ˆz ′ | determines the magnitude of E(r), |r − z we may contributions, yielding: use the approximation: Z +L/2 e e r; z ′ ) ˆz ′ | ≈ r (magnitude) E(r) ≈ dE(ˆ (9.88) |r − z (9.93) −L/2

Expanding this expression: ηβ e E(r) ≈j 4π

Z

+L/2 −L/2

e ′ ) (sin θ′ ) e θˆ′ I(z

−jβ |r−ˆ zz ′ |

dz ′ ˆz ′ | |r − z (9.89)

ˆz ′ | determines phase, we have to be Insofar as |r − z more careful. The part of the integrand of Equation 9.90 that exhibits varying phase is ′ e−jβ |r−ˆzz | . Using Equation 9.91, we find e−jβ |r−ˆzz | ≈ e−jβr e+jβz ′



cos θ

(9.94)

9.7. FAR-FIELD RADIATION FROM A HALF-WAVE DIPOLE These simplifications are known collectively as a far field approximation, since they are valid only for distances “far” from the source. Applying these simplifications for magnitude and phase to Equation 9.90, we obtain: −jβr

ˆ ηβ e e E(r) ≈ θj (sin θ) 4π r Z +L/2 e ′ )e+jβz′ cos θ dz ′ · I(z

(9.95)

−L/2

Finally, it is common to eliminate the factor of β in the magnitude using the relationship β = 2π/λ, yielding: −jβr

ˆ ηe e E(r) ≈ θj (sin θ) " 2Z r # 1 +L/2 e ′ +jβz′ cos θ ′ · I(z )e dz λ −L/2

(9.96)

ˆThe electric field radiated by a thin, straight, z directed current filament of length L located at the origin and aligned along the z axis is given by Equation 9.96. This expression is valid for r ≫ L and r ≫ λ. At field points satisfying the conditions r ≫ L and r ≫ λ, the wave appears to be locally planar. Therefore, we are justified using the plane wave e = (1/η)ˆr × E e to calculate H. e relationship H

161

9.7 Far-Field Radiation from a Half-Wave Dipole [m0200]

A simple and important current distribution is that of the thin half-wave dipole (HWD), shown in Figure 9.16. This is the distribution expected on a thin straight wire having length L = λ/2, where λ is wavelength. This distribution is described mathematically as follows:   L e ≈ I0 cos π z I(z) for |z| ≤ (9.97) L 2 where I0 (SI base units of A) is a complex-valued constant indicating the maximum magnitude of the current and its phase. Note that the current is zero at e = 0 for |z| = L/2. the ends of the dipole; i.e., I(z) Note also that this “cosine pulse” distribution is very similar to the triangular distribution of the ESD, and is reminiscent of the sinusoidal variation of current in a standing wave. Since L = λ/2 for the HWD, Equation 9.97 may equivalently be written:   e ≈ I0 cos 2π z I(z) (9.98) λ The electromagnetic field radiated by this distribution

As a check, one may readily verify that Equation 9.96 yields the expected result for the electrically-short dipole (Section 9.5).

c C. Wang CC BY-SA 4.0

Figure 9.16: Current distribution of the half-wave dipole (HWD).

CHAPTER 9. RADIATION

162 of current may be calculated using the method described in Section 9.6, in particular: −jβr

ˆ ηe e E(r) ≈ θj (sin θ) # " 2Z r 1 +L/2 e ′ +jβz′ cos θ ′ I(z )e dz · λ −L/2

(9.99)

which is valid for field points r far from the dipole; i.e., for r ≫ L and r ≫ λ. For the HWD, the quantity in square brackets is   Z ′ I0 +λ/4 z′ e+jβz cos θ dz ′ cos 2π (9.100) λ −λ/4 λ The evaluation of this integral is straightforward, but tedious. The integral reduces to I0 cos [(π/2) cos θ] π sin2 θ Substitution into Equation 9.99 yields −jβr

(9.101) Figure 9.17: Comparison of the magnitude of the radiated field of the HWD to that of an electrically-short dipole also oriented along the z-axis. This result is for any radial plane that includes the z-axis. (9.102)

ˆ ηI0 cos [(π/2) cos θ] e e E(r) ≈ θj 2π sin θ r The magnetic field may be determined from this result using Ampere’s law. However, a simpler method is to use the fact that the electric field, magnetic field, and direction of propagation ˆr are mutually perpendicular and related by: e = 1 ˆr × E e H η

c C. Wang CC BY-SA 4.0

(9.103)

9.8 Radiation from Surface and Volume Distributions of Current [m0221]

This relationship indicates that the magnetic field will ˆ be +φ-directed.

In Section 9.3, a solution was developed for radiation from current located at a single point r′ . This current distribution was expressed mathematically as a The magnitude and polarization of the radiated field is current moment, as follows: similar to that of the electrically-short dipole (ESD; e Section 9.5). A comparison of the magnitudes in any J(r) = ˆl Ie ∆l δ(r − r′ ) (9.104) radial plane containing the z-axis is shown in where ˆl is the direction of current flow, Ie has units of Figure 9.17. For either current distribution, the maximum magnitude of the fields occurs in the z = 0 current (SI base units of A), ∆l has units of length (SI base units of m), and δ(r) is the volumetric sampling plane. For a given terminal current I0 , the maximum function (Dirac “delta” function; SI base units of magnitude is greater for the HWD than for the ESD. e has SI base units of Both current distributions yield zero magnitude along m−3 ). In this formulation, J 2 A/m . The magnetic vector potential radiated by this the axis of the dipole. The polarization characteristics current, observed at the field point r, was found to be of the fields of both current distributions are identical. Additional Reading: • “Dipole antenna” (section entitled “Half-wave dipole”) on Wikipedia.

e−γ |r−r | e A(r) = ˆl µ Ie ∆l 4π |r − r′ | ′

(9.105)

This solution was subsequently generalized to obtain the radiation from any distribution of line current; i.e.,

9.8. RADIATION FROM SURFACE AND VOLUME DISTRIBUTIONS OF CURRENT any distribution of current that is constrained to flow along a single path through space, as along an infinitesimally-thin wire. In this section, we derive an expression for the radiation from current that is constrained to flow along a surface and from current which flows through a volume. The solution in both cases can be obtained by “recycling” the solution for line current as follows. Letting ∆l shrink to the differential length dl, Equation 9.104 becomes: e dJ(r) = ˆl Ie dl δ(r − r′ )

(9.106)

Subsequently, Equation 9.105 becomes: −γ |r−r′ |

e r′ ) = ˆl µ Ie dl e dA(r; 4π |r − r′ |

(9.107)

163

of m2 ). To emphasize that this is precisely the same current moment, note that Jes ds, like Ie dl, has units of A·m. Similarly, e dJ(r) = ˆl Je dv δ(r − r′ )

(9.110)

where Je has units of volume current density (SI base units of A/m2 ) and dv has units of volume (SI base units of m3 ). Again, Je dv has units of A·m. Summarizing, we have found that Ie dl = Jes ds = Je dv

(9.111)

all describe the same differential current moment. Thus, we may obtain solutions for surface and volume distributions of current simply by replacing Ie dl in Equation 9.107, and subsequently in Equation 9.108, with the appropriate quantity from Equation 9.111. For a surface current distribution, we obtain:

e r′ ) is used to denote the where the notation dA(r; magnetic vector potential at the field point r due to Z −γ |r−r′ | µ the differential-length current moment at the source ′ e e e ds (9.112) Js (r ) A(r) = 4π S |r − r′ | point r′ . Next, consider that any distribution of current can be described as a distribution of current es (r′ ) , ˆl(r′ )Jes (r′ ) and where S is the moments. By the principle of superposition, the where J radiation from this distribution of current moments surface over which the current flows. Similarly for a can be calculated as the sum of the radiation from the volume current distribution, we obtain: individual current moments. This is expressed Z mathematically as follows: −γ |r−r′ | µ ′ e e e dv A(r) = J(r ) (9.113) Z 4π V |r − r′ | ′ e e r) A(r) = dA(r; (9.108) where the integral is over r′ ; i.e., summing over the source current.

e ′ ) , ˆl(r′ )J(r e ′ ) and where V is the volume where J(r in which the current flows.

The magnetic vector potential corresponding to If we substitute Equation 9.107 into Equation 9.108, radiation from a surface and volume distribution we obtain the solution derived in Section 9.3, which is of current is given by Equations 9.112 and 9.113, specific to line distributions of current. To obtain the respectively. solution for surface and volume distributions of current, we reinterpret the definition of the differential e current moment in Equation 9.106. Note that this Given A(r), the magnetic and electric fields may be current is completely specified by its direction (ˆl) and determined using the procedure developed in the quantity Ie dl, which has SI base units of A·m. We Section 9.2. may describe the same current distribution [m0217] alternatively as follows: e dJ(r) = ˆl Jes ds δ(r − r′ )

(9.109)

where Jes has units of surface current density (SI base units of A/m) and ds has units of area (SI base units

164

CHAPTER 9. RADIATION

Image Credits c Sevenchw (C. Wang), Fig. 9.1: https://commons.wikimedia.org/wiki/File:A z-directed current moment located at the origin.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 9.2: https://commons.wikimedia.org/wiki/File:A z-directed current moment located at the origin.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 9.3: https://commons.wikimedia.org/wiki/File:Current moment displaced from the origin.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 9.4: https://commons.wikimedia.org/wiki/File:A filament of current lying along the path c.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 9.5: https://commons.wikimedia.org/wiki/File:A z-directed current moment located at the origin.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 9.6: https://commons.wikimedia.org/wiki/File:Current distribution of the esd.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 9.7: https://commons.wikimedia.org/wiki/File:Esd approximated as a large number of hertzian dipoles.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 9.8: https://commons.wikimedia.org/wiki/File:Parallel ray approximation for an esd.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), Fig. 9.9: https://commons.wikimedia.org/wiki/File:Magnitude of the Radiated Field.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), Fig. 9.10: https://commons.wikimedia.org/wiki/File:Electric and Magnetic Fields in Plane of Constant Theta.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), https://commons.wikimedia.org/wiki/File:FHplaneMag.svg, Fig. 9.11: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), https://commons.wikimedia.org/wiki/File:FHplanePol.svg, Fig. 9.12: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 9.13: https://commons.wikimedia.org/wiki/File:A thin straight distribution of radiating current.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 9.14: https://commons.wikimedia.org/wiki/File:Esd approximated as a large number of hertzian dipoles.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/).

9.8. RADIATION FROM SURFACE AND VOLUME DISTRIBUTIONS OF CURRENT c Sevenchw (C. Wang), Fig. 9.15: https://commons.wikimedia.org/wiki/File:Parallel ray approximation for an esd.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 9.16: https://commons.wikimedia.org/wiki/File:Current distribution of the hwd.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 9.17: https://commons.wikimedia.org/wiki/File:Comparison radiated field magnitude of hwd to esd.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/).

165

Chapter 10

Antennas 10.1 How Antennas Radiate [m0201]

An antenna is a transducer; that is, a device which converts signals in one form into another form. In the case of an antenna, these two forms are (1) conductor-bound voltage and current signals and (2) electromagnetic waves. Traditional passive antennas are capable of this conversion in either direction. In this section, we consider the transmit case, in which a conductor-bound signal is converted into a radiating electromagnetic wave. Radiation from an antenna is due to the time-varying current that is excited by the bound electrical signal applied to the antenna terminals.

applied to the input of an ideal twin lead transmission line. The spacing between conductors is much less than a wavelength, and the output of the transmission line is terminated into an open circuit. Without any additional information, we already know two things about the current on this transmission line. First, we know the current must be identically zero at the end of the transmission line. Second, we know the current on the two conductors comprising the transmission line must be related as indicated in Figure 10.1. That is, at any given position on the transmission line, the current on each conductor is equal in magnitude and flows in opposite directions.

Further note that Figure 10.1 depicts only the situation over an interval of one-half of the period of the sinusoidal source. For the other half-period, the Why ideal transmission lines don’t radiate. To direction of current will be in the direction opposite describe the process that allows an antenna to that depicted in Figure 10.1. In other words: The transmit, it is useful to first consider the scenario source is varying periodically, so the sign of the depicted in Figure 10.1. Here a sinusoidal source is current is changing every half-period. Generally, time-varying currents give rise to radiation. Therefore, we should expect the currents depicted in Figure 10.1 might radiate. We can estimate the radiation from the transmission line by interpreting the current as a collection of Hertzian dipoles. For a review of Hertzian dipoles, see Section 9.4; however, all we need to know to address the present problem is that superposition applies. That is, the total radiation is the sum of the radiation from the individual Hertzian dipoles. Once again looking back at Figure 10.1, note that each Hertzian dipole representing current on one conductor has an c S. Lally CC BY-SA 4.0

associated Hertzian dipole representing the current on Figure 10.1: Twin lead transmission line terminated the other conductor, and is only a tiny fraction of a wavelength distant. Furthermore, these pairs of into an open circuit, giving rise to a standing wave.

c 2020 S.W. Ellingson CC BY SA 4.0. https://doi.org/10.21061/electromagnetics-vol-2 Electromagnetics Vol. 2.

10.1. HOW ANTENNAS RADIATE Hertzian dipoles are identical in magnitude but opposite in sign. Therefore, the radiated field from any such pair of Hertzian dipoles is approximately zero at distances sufficiently far from the transmission line. Continuing to sum all such pairs of Hertzian dipoles, the radiated field remains approximately zero at distances sufficiently far from the transmission line. We come to the following remarkable conclusion: The radiation from an ideal twin lead transmission line with open circuit termination is negligible at distances much greater than the separation between the conductors. Before proceeding, let us be clear about one thing: The situation at distances which are not large relative to separation between the conductors is quite different. This is because the Hertzian dipole pairs do not appear to be quite so precisely collocated at field points close to the transmission line. Subsequently the cancellation of fields from Hertzian dipole pairs is less precise. The resulting sum fields are not negligible and depend on the separation between the conductors. For the present discussion, it suffices to restrict our attention to the simpler “far field” case. A simple antenna formed by modifying twin lead transmission line. Let us now consider a modification to the previous scenario, shown in Figure 10.2. In the new scenario, the ends of the twin lead are bent into right angles. This new section has an overall length which is much less than one-half wavelength. To determine the current distribution on the modified section, we first note that the current must still be precisely zero at the ends of the conductors, but not necessarily at the point at which

c S. Lally CC BY-SA 4.0

Figure 10.2: End of the twin lead transmission line fashioned into an electrically-short dipole (ESD).

167 they are bent. Since the modified section is much shorter than one-half wavelength, the current distribution on this section must be very simple. In fact, the current distribution must be that of the electrically-short dipole (ESD), exhibiting magnitude which is maximum at the center and decreasing approximately linearly to zero at the ends. (See Section 9.5 for a review of the ESD.) Finally, we note that the current should be continuous at the junction between the unmodified and modified sections. Having established that the modified section exhibits the current distribution of an ESD, and having previously determined that the unmodified section of transmission line does not radiate, we conclude that this system radiates precisely as an ESD does. Note also that we need not interpret the ESD portion of the transmission line as a modification of the transmission line: Instead, we may view this system as an unmodified transmission line attached to an antenna, which in this case in an ESD. The general case. Although we developed this insight for the ESD specifically, the principle applies generally. That is, any antenna – not just dipoles – can be viewed as a structure that can support a current distribution that radiates. Elaborating: A transmitting antenna is a device that, when driven by an appropriate source, supports a timevarying distribution of current resulting in an electromagnetic wave that radiates away from the device. Although this may seem to be simply a restatement of the definition at the beginning of this section – and it is – we now see how this can happen, and also why transmission lines typically aren’t also antennas.

CHAPTER 10. ANTENNAS

168

10.2 Power Radiated by an Electrically-Short Dipole

integrated over an area (m2 ) gives power (W). Anticipating that this problem will be addressed in spherical coordinates, we note that ds = ˆrr2 sin θ dθ dφ

[m0207]

In this section, we determine the total power radiated by an electrically-short dipole (ESD) antenna in response to a sinusoidally-varying current applied to the antenna terminals. This result is both useful on its own and necessary as an intermediate result in determining the impedance of the ESD. The ESD is introduced in Section 9.5, and a review of that section is suggested before attempting this section. In Section 9.5, it is shown that the electric field ˆ-oriented ESD located at intensity in the far field of a z the origin is −jβr ˆ I0 · βL (sin θ) e e (10.1) E(r) ≈ θjη 8π r where r is the field point, I0 is a complex number representing the peak magnitude and phase of the sinusoidally-varying terminal current, L is the length of the ESD, and β is the phase propagation constant 2π/λ where λ is wavelength. Note that L ≪ λ since this is an ESD. Also note that Equation 10.1 is valid only for the far-field conditions r ≫ L and r ≫ λ, and presumes propagation in simple (linear, homogeneous, time-invariant, isotropic) media with negligible loss.

Given that we have already limited scope to the far field, it is reasonable to approximate the electromagnetic field at each field point r as a plane wave propagating radially away from the antenna; i.e., in the ˆr direction. Under this assumption, the time-average power density is e 2 E(r) S(r) = ˆr (10.2) 2η

(10.4)

and subsequently:

Prad =

=

Z

π θ=0

1 2η

Z

Z

  e 2 E(r)    2 ˆr  · ˆrr sin θ dθ dφ 2η φ=0 2π

π

θ=0

Z



φ=0

e 2 2 E(r) r sin θ dθ dφ (10.5)

Returning to Equation 10.1, we note:

2 2 |I0 | (βL) e 2 2 1 (sin θ) 2 E(r) ≈ η 2 64π 2 r

(10.6)

Substitution into Equation 10.5 yields: 2

Prad ≈ η

|I0 | (βL) 128π 2

2

Z

π θ=0

Z



sin3 θ dθ dφ (10.7)

φ=0

Note that this may be factored into separate integrals over θ and φ. The integral over φ is simply 2π, leaving: 2

Prad ≈ η

|I0 | (βL) 64π

2

Z

π

sin3 θ dθ

(10.8)

θ=0

The remaining integral is equal to 4/3, leaving: 2

Prad ≈ η

|I0 | (βL) 48π

2

(10.9)

This completes the derivation, but it is useful to check units. Recall that β has SI base units of rad/m, so βL 2 has units of radians. This leaves η |I0 | , which has SI 2 base units of Ω · A = W, as expected.

The power radiated by an ESD in response to the current I0 applied at the terminals is given by where η is the wave impedance of the medium. The Equation 10.9. total power Prad radiated by the antenna is simply S(r) integrated over any closed surface S that encloses the antenna. Thus: I Finally, it is useful to consider how various Prad = S(r) · ds (10.3) parameters affect the radiated power. First, note that S the radiated power is proportional to the square of the where ds is the outward-facing differential element of terminal current. Second, note that the product surface area. In other words, power density (W/m2 ) βL = 2πL/λ is the electrical length of the antenna;

10.3. POWER DISSIPATED BY AN ELECTRICALLY-SHORT DIPOLE that is, the length of the antenna expressed in radians, where 2π radians is one wavelength. Thus, we see that the power radiated by the antenna increases as the square of electrical length.

169

10.3 Power Dissipated by an Electrically-Short Dipole [m0208]

Example 10.1. Power radiated by an ESD. A dipole is 10 cm in length and is surrounded by free space. A sinusoidal current having frequency 30 MHz and peak magnitude 100 mA is applied to the antenna terminals. What is the power radiated by the antenna? Solution. If no power is dissipated within the antenna, then all power is radiated. The wavelength λ = c/f ∼ = 10 m, so L ∼ = 0.01λ. This certainly qualifies as electrically-short, so we may use Equation 10.9. In the present problem, η ∼ = 376.7 Ω (the free space wave impedance), I0 = 100 mA, and β = 2π/λ ∼ = 0.628 rad/m. Thus, we find that the radiated power is ≈ 98.6 µW.

The power delivered to an antenna by a source connected to the terminals is nominally radiated. However, it is possible that some fraction of the power delivered by the source will be dissipated within the antenna. In this section, we consider the dissipation due to the finite conductivity of materials comprising the antenna. Specifically, we determine the total power dissipated by an electrically-short dipole (ESD) antenna in response to a sinusoidally-varying current applied to the antenna terminals. (The ESD is introduced in Section 9.5, and a review of that section is suggested before attempting this section.) The result allows us to determine radiation efficiency and is a step toward determining the impedance of the ESD. Consider an ESD centered at the origin and aligned along the z axis. In Section 9.5, it is shown that the current distribution is:   e ≈ I0 1 − 2 |z| I(z) (10.10) L

where I0 (SI base units of A) is a complex-valued constant indicating the maximum current magnitude and phase, and L is the length of the ESD. This current distribution may be interpreted as a set of discrete very-short segments of constant-magnitude current (sometimes referred to as “Hertzian dipoles”). In this interpretation, the nth current segment is e n ). located at z = zn and has magnitude I(z

Now let us assume that the material comprising the ESD is a homogeneous good conductor. Then each segment exhibits the same resistance Rseg . Subsequently, the power dissipated in the nth segment is 2 1 e Pseg (zn ) = I(z (10.11) n ) Rseg 2 The segment resistance can be determined as follows. Let us assume that the wire comprising the ESD has circular cross-section of radius a. Since the wire is a good conductor, the segment resistance may be

CHAPTER 10. ANTENNAS

170

Now let us return to interpretation of the ESD as a circuit component. We have found that applying the current I0 to the terminals results in the dissipated (10.12) power indicated in Equation 10.18. A current source driving the ESD does not perceive the current distribution of the ESD nor does it perceive the where µ is permeability, f is frequency, σ is varying power over the length of the ESD. Instead, conductivity, and ∆l is the length of the segment. the current source perceives only a net resistance Substitution into Equation 10.11 yields: Rloss such that r 2 1 µf e 1 2 Pseg (zn ) ≈ (10.13) I(zn ) ∆l (10.19) Ploss = |I0 | Rloss 4a πσ 2

calculated using Equation 4.17 (Section 4.2, “Impedance of a Wire”): r 1 µf ∆l · Rseg ≈ 2 πσ a

Now the total power dissipated in the antenna, Ploss , can be expressed as the sum of the power dissipated in each segment: # " r N 2 X µf e 1 (10.14) Ploss ≈ I(zn ) ∆l 4a πσ n=1 where N is the number of segments. This may be rewritten as follows: r N 2 µf X e 1 (10.15) Ploss ≈ I(zn ) ∆l 4a πσ n=1

Comparing Equations 10.18 and 10.19, we find r L µf Rloss ≈ (10.20) 6a πσ The power dissipated within an ESD in response to a sinusoidal current I0 applied at the terminals 2 is 12 |I0 | Rloss where Rloss (Equation 10.20) is the resistance perceived by a source applied to the ESD’s terminals.

Note that Rloss is not the impedance of the ESD. Rloss is merely the contribution of internal loss to the Reducing ∆l to the differential length dz , we may impedance of the ESD. The impedance of the ESD write this in the following integral form: must also account for contributions from radiation r Z resistance (an additional real-valued contribution) and 1 µf +L/2 e ′ 2 ′ Ploss ≈ (10.16) energy storage (perceived as a reactance). These I(z ) dz 4a πσ z′ =−L/2 quantities are addressed in other sections of this book. It is worth noting that the above expression applies to any straight wire antenna of length L. For the ESD Example 10.2. Power dissipated within an specifically, the current distribution is given by ESD. Equation 10.10. Making the substitution: r   2 Z A dipole is 10 cm in length, 1 mm in radius, and 1 µf +L/2 2 ′ Ploss ≈ I0 1 − |z | dz ′ is surrounded by free space. The antenna is 4a πσ z′ =−L/2 L comprised of aluminum having conductivity r 2 Z +L/2 µf 2 1 2 ′ ′ ≈ 3.7 × 107 S/m and µ ≈ µ0 . A sinusoidal 1 − |z | dz |I0 | ≈ 4a πσ L ′ current having frequency 30 MHz and peak z =−L/2 magnitude 100 mA is applied to the antenna (10.17) terminals. What is the power dissipated within The integral is straightforward to solve, albeit a bit this antenna? tedious. The integral is found to be equal to L/3, so we obtain: Solution. The wavelength λ = c/f ∼ = 10 m, so r ∼ L = 10 cm 0.01λ. This certainly qualifies as = 1 µf 2 L |I0 | · Ploss ≈ electrically-short, so we may use 4a πσ 3 r Equation 10.20. In the present problem, L µf 2 a = 1 mm and σ ≈ 3.7 × 107 S/m. Thus, we ≈ |I0 | (10.18) 12a πσ ′

10.4. REACTANCE OF THE ELECTRICALLY-SHORT DIPOLE find that the loss resistance Rloss ≈ 9.49 mΩ. Subsequently, the power dissipated within this antenna is Ploss =

1 2 |I0 | Rloss ≈ 47.5 µW 2

(10.21)

We conclude this section with one additional caveat: Whereas this section focuses on the limited conductivity of wire, other physical mechanisms may contribute to the loss resistance of the ESD. In particular, materials used to coat the antenna or to provide mechanical support near the terminals may potentially absorb and dissipate power that might otherwise be radiated.

171

10.4 Reactance of the Electrically-Short Dipole [m0210]

For any given time-varying voltage appearing across the terminals of an electrically-short antenna, there will be a time-varying current that flows in response. The ratio of voltage to current is impedance. Just as a resistor, capacitor, or inductor may be characterized in terms of an impedance, any antenna may be characterized in terms of this impedance. The real-valued component of this impedance accounts for power which is radiated away from the antenna (Section 10.2) and dissipated within the antenna (Section 10.3). The imaginary component of this impedance – i.e., the reactance – typically represents energy storage within the antenna, in the same way that the reactance of a capacitor or inductor represents storage of electrical or magnetic energy, respectively. In this section, we determine the reactance of the electrically-short dipole (ESD). Reactance of a zero-length dipole. We begin with what might seem initially to be an absurd notion: A dipole antenna having zero length. However, such a dipole is certainly electrically-short, and in fact serves as an extreme condition from which we can deduce some useful information. Consider Figure 10.3, which shows a zero-length dipole being driven by a source via a transmission line. It is immediately apparent that such a dipole is equivalent to an open circuit. So, the impedance of a zero-length dipole is equal to that of an open circuit.

c S. Lally CC BY-SA 4.0

Figure 10.3: Transmission line terminated into a zerolength dipole, which is equivalent to an open circuit.

CHAPTER 10. ANTENNAS

172 What is the impedance of an open circuit? One might be tempted to say “infinite,” since the current is zero independently of the voltage. However, we must now be careful to properly recognize the real and imaginary components of this infinite number, and signs of these components. In fact, the impedance of an open circuit is 0 − j∞. One way to confirm this is via basic transmission line theory; i.e., the input impedance of transmission line stubs. A more direct justification is as follows: The real part of the impedance is zero because there is no transfer of power into the termination. The magnitude of the imaginary part of the impedance must be infinite because the current is zero. The sign of the imaginary component is negative because the reflected current experiences a sign change (required to make the total current zero), whereas the reflected voltage does not experience a sign change (and so is not canceled at the terminals). Thus, the impedance of an open circuit, and the zero-length dipole, is 0 − j∞. Reactance of a nearly-zero-length dipole. Let us now consider what happens as the length of the dipole increases from zero. Since such a dipole is short compared to a wavelength, the variation with length must be very simple. The reactance remains large and negative, but can only increase (become less negative) monotonically with increasing electrical length. This trend continues until the dipole is no longer electrically short. Thus, we conclude that the reactance of an ESD is always large and negative. The reactance of an ESD is very large and negative, approaching the reactance of an open circuit termination as length decreases to zero.

(see, e.g., Johnson (1993) in “Additional References” at the end of this section):     L 120 Ω ln −1 (10.22) XA ≈ − πL/λ 2a where L is length and a ≪ L is the radius of the wire comprising the ESD. Note that this expression yields the expected behavior; i.e., XA → −∞ as L → 0, and increases monotonically as L increases from zero. Example 10.3. Reactance of an ESD. A dipole is 10 cm in length, 1 mm in radius, and is surrounded by free space. What is the reactance of this antenna at 30 MHz? Solution. The wavelength λ = c/f ∼ = 10 m, so L = 10 cm ∼ = 0.01λ. This certainly qualifies as electrically-short, so we may use Equation 10.22. Given a = 1 mm, we find the reactance XA ≈ −11.1 kΩ. For what it’s worth, this antenna exhibits approximately the same reactance as a 0.47 pF capacitor at the same frequency. The reactance of the ESD is typically orders of magnitude larger than the real part of the impedance of the ESD. The reactance of the ESD is also typically very large compared to the characteristic impedance of typical transmission lines (usually 10s to 100s of ohms). This makes ESDs quite difficult to use in practical transmit applications. Additional Reading:

Note that this behavior is similar to that of a capacitor, whose impedance is also negative and increases toward zero with increasing frequency. Thus, the reactance of an ESD is sometimes represented in circuit diagrams as a capacitor. However, the specific dependence of reactance on frequency is different from that of a capacitor (as we shall see in a moment), so this model is generally valid only for analysis at one frequency at time. An approximate expression for the reactance of an ESD. Sometimes it is useful to be able to estimate the reactance XA of an ESD. Although a derivation is beyond the scope of this text, a suitable expression is

• R.C. Johnson (Ed.), Antenna Systems Handbook (Ch. 4), McGraw-Hill, 1993.

10.5. EQUIVALENT CIRCUIT MODEL FOR TRANSMISSION; RADIATION EFFICIENCY

10.5 Equivalent Circuit Model for Transmission; Radiation Efficiency [m0202]

A radio transmitter consists of a source which generates the electrical signal intended for transmission, and an antenna which converts this signal into a propagating electromagnetic wave. Since the transmitter is an electrical system, it is useful to be able to model the antenna as an equivalent circuit. From a circuit analysis point of view, it should be possible to describe the antenna as a passive one-port circuit that presents an impedance to the source. Thus, we have the following question: What is the equivalent circuit for an antenna which is transmitting? We begin by emphasizing that the antenna is passive. That is, the antenna does not add power. Invoking the principle of conservation of power, there are only three possible things that can happen to power that is delivered to the antenna by the transmitter:1 • Power can be converted to a propagating electromagnetic wave. (The desired outcome.) • Power can be dissipated within the antenna.

173

In the phasor domain, ZA is a complex-valued quantity and therefore has, in general, a real-valued component and an imaginary component. We may identify those components using the power conservation argument made previously: Since the real-valued component must represent power transfer and the imaginary component must represent energy storage, we infer: ZA , RA + jXA

(10.24)

where RA represents power transferred to the antenna, and XA represents energy stored by the antenna. Note that the energy stored by the antenna is being addressed in precisely the same manner that we address energy storage in a capacitor or an inductor; in all cases, as reactance. Further, we note that RA consists of components Rrad and Rloss as follows: ZA = Rrad + Rloss + jXA

(10.25)

where Rrad represents power transferred to the antenna and subsequently radiated, and Rloss represents power transferred to the antenna and subsequently dissipated. To confirm that this model works as expected, consider what happens when a voltage is applied across the antenna terminals. The current IeA flows and the time-average power PA transferred to the

• Energy can be stored by the antenna, analogous to the storage of energy in a capacitor or inductor. We also note that these outcomes can occur in any combination. Taking this into account, we model the antenna using the equivalent circuit shown in Figure 10.4. Since the antenna is passive, it is reasonable to describe it as an impedance ZA which is (by definition) the ratio of voltage VeA to current IeA at the terminals; i.e., ZA , 1 Note

VeA IeA

(10.23)

that “delivered” power means power accepted by the antenna. We are not yet considering power reflected from the antenna due to impedance mismatch.

c S. Lally CC BY-SA 4.0

Figure 10.4: Equivalent circuit for an antenna which is transmitting.

CHAPTER 10. ANTENNAS

174 antenna is

1 n e e∗ o Re VA IA (10.26) 2 where we have assumed peak (as opposed to root mean squared) units for voltage and current. Since VeA = ZA IeA , we have: o 1 n ∗ PA = Re (Rrad + Rloss + jXA ) IeA IeA 2 (10.27) which reduces to: 1 2 1 2 PA = IeA Rrad + IeA Rloss (10.28) 2 2 PA =

Radiation resistance. Rrad is referred to as radiation resistance. Equation 10.29 tells us that −2 Rrad = 2Prad IeA

(10.32)

This equation suggests the following procedure: We apply current IeA to the antenna terminals, and then determine the total power Prad radiated from the antenna in response. For an example of this procedure, see Section 10.2 (“Total Power Radiated by an Electrically-Short Dipole”). Given IeA and Prad , one may then use Equation 10.32 to determine Rrad .

Loss resistance. Loss resistance represents the dissipation of power within the antenna, which is usually attributable to loss intrinsic to materials comprising or surrounding the antenna. In many cases, antennas are made from good conductors – metals, in particular – so that Rloss is very low compared to Rrad . For such antennas, loss is often so 1 e 2 (10.30) low compared to Rrad that Rloss may be neglected. Ploss , IA Rloss 2 In the case of the electrically-short dipole, Rloss is typically very small but Rrad is also very small, so representing power dissipated within the antenna. both must be considered. In many other cases, e antennas contain materials with substantially greater The reactance XA will play a role in determining IA e loss than metal. For example, a microstrip patch given VA (and vice versa), but does not by itself antenna implemented on a printed circuit board account for disposition of power. Again, this is exactly analogous to the role played by inductors and typically has non-negligible Rloss because the dielectric material comprising the antenna exhibits capacitors in a circuit. significant loss. The utility of this equivalent circuit formalism is that it allows us to treat the antenna in the same manner as Antenna reactance. The reactance term jXA any other component, and thereby facilitates analysis accounts for energy stored by the antenna. This may be due to reflections internal to the antenna, or due to using conventional electric circuit theory and transmission line theory. For example: Given ZA , we energy associated with non-propagating electric and know how to specify the output impedance ZS of the magnetic fields surrounding the antenna. The presence of significant reactance (i.e., |XA | transmitter so as to minimize reflection from the comparable to or greater than |RA |) complicates antenna: We would choose ZS = ZA , since in this efforts to establish the desired impedance match to case the voltage reflection coefficient would be the source. For an example, see Section 10.4 ZA − ZS (“Reactance of the Electrically-Short Dipole”). Γ= =0 (10.31) ZA + ZS Radiation efficiency. When Rloss is non-negligible, Alternatively, we might specify ZS so as to maximize it is useful to characterize antennas in terms of their power transfer to the antenna: We would choose radiation efficiency erad , defined as the fraction of ∗ ZS = ZA ; i.e., conjugate matching. power which is radiated compared to the total power delivered to the antenna; i.e., In order to take full advantage of this formalism, we require values for Rrad , Rloss , and XA . These Prad (10.33) erad , quantities are considered below. PA As expected, the power transferred to the antenna is the sum of 1 2 (10.29) Prad , IeA Rrad 2 representing power transferred to the radiating electromagnetic field, and

10.6. IMPEDANCE OF THE ELECTRICALLY-SHORT DIPOLE

10.6 Impedance of the Electrically-Short Dipole

Using Equations 10.28–10.30, we see that this efficiency can be expressed as follows: erad =

Rrad Rrad + Rloss

(10.34) [m0204]

Once again, the equivalent circuit formalism proves useful. Example 10.4. Impedance of an antenna. The total power radiated by an antenna is 60 mW when 20 mA (rms) is applied to the antenna terminals. The radiation efficiency of the antenna is known to be 70%. It is observed that voltage and current are in-phase at the antenna terminals. Determine (a) the radiation resistance, (b) the loss resistance, and (c) the impedance of the antenna. Solution. From the statement, problem e Prad = 60 mW, IA = 20 mA (rms), and erad = 0.7. Also, the fact that voltage and current are in-phase at the antenna terminals indicates that XA = 0. From Equation 10.32, the radiation resistance is Rrad

2 · (60 mW) ≈ √ = 150 Ω 2 · 20 mA 2

(10.35)

In this section, we determine the impedance of the electrically-short dipole (ESD) antenna. The physical theory of operation for the ESD is introduced in Section 9.5, and additional relevant aspects of the ESD antenna are addressed in Sections 10.1–10.4. The concept of antenna impedance is addressed in Section 10.5. Review of those sections is suggested before attempting this section. The impedance of any antenna may be expressed as ZA = Rrad + Rloss + jXA

1 − erad Rrad ∼ = 64.3 Ω erad

Radiation resistance. The radiation resistance of any antenna can be expressed as: Rrad = 2Prad |I0 |

Since ZA = Rrad + Rloss + jXA , we find ZA ∼ = 214.3 + j0 Ω. This will be the ratio of voltage to current at the antenna terminals regardless of the source current.

−2

(10.38)

where |I0 | is the magnitude of the current at the antenna terminals, and Prad is the resulting total power radiated. For an ESD (Section 10.2): 2

Prad ≈ η

|I0 | (βL) 48π

so (10.36)

(10.37)

where the real-valued quantities Rrad , Rloss , and XA represent the radiation resistance, loss resistance, and reactance of the antenna, respectively.

Solving Equation 10.34 for the loss resistance, we find: Rloss =

175

2

(10.39)

2

(βL) (10.40) 24π It is useful to have an alternative form of this expression in terms of wavelength λ. This is derived as follows. First, note: Rrad ≈ η

βL =

L 2π L = 2π λ λ

(10.41)

where L/λ is the antenna length in units of wavelength. Substituting this expression into Equation 10.40:  2 L 1 Rrad ≈ η 2π λ 24π  2 π L ≈η (10.42) 6 λ

CHAPTER 10. ANTENNAS

176 Assuming free-space conditions, η ∼ = 376.7 Ω, which is ≈ 120π Ω. Subsequently, Rrad ≈ 20π 2

 2 L λ

where a ≪ L is assumed. Example 10.5. Impedance of an ESD.

(10.43)

This remarkably simple expression indicates that the radiation resistance of an ESD is very small (since L ≪ λ for an ESD), but increases as the square of the length. At this point, a warning is in order. The radiation resistance of an “electrically-short dipole” is 2 sometimes said to be 80π 2 (L/λ) ; i.e., 4 times the right side of Equation 10.43. This higher value is not for a physically-realizable ESD, but rather for the Hertzian dipole (sometimes also referred to as an “ideal dipole”). The Hertzian dipole is an electrically-short dipole with a current distribution that has uniform magnitude over the length of the dipole.2 The Hertzian dipole is quite difficult to realize in practice, and nearly all practical electrically-short dipoles exhibit a current distribution that is closer to that of the ESD. The ESD has a current distribution which is maximum in the center and goes approximately linearly to zero at the ends. The factor-of-4 difference in the radiation resistance of the Hertzian dipole relative to the (practical) ESD is a straightforward consequence of the difference in the current distributions.

A thin, straight dipole antenna operates at 30 MHz in free space. The length and radius of the dipole are 1 m and 1 mm respectively. The dipole is comprised of aluminum having conductivity ≈ 3.7 × 107 S/m and µ ≈ µ0 . What is the impedance and radiation efficiency of this antenna? Solution. The free-space wavelength at f = 30 MHz is λ = c/f ∼ = 10 m. Therefore, L∼ = 0.1λ, and this is an ESD. Since this is an ESD, we may compute the radiation resistance using Equation 10.43, yielding Rrad ≈ 1.97 Ω. The radius a = 1 mm and σ ≈ 3.7 × 107 S/m; thus, we find that the loss resistance Rloss ≈ 94.9 mΩ using Equation 10.44. Using Equation 10.45, the reactance is found to be XA ≈ −1991.8 Ω. The impedance of this antenna is therefore ZA = Rrad + Rloss + jXA ≈ 2.1 − j1991.8 Ω

(10.46)

and the radiation efficiency of this antenna is erad =

Rrad ≈ 95.4% Rrad + Rloss

(10.47)

Loss resistance. The loss resistance of the ESD is derived in Section 10.3 and is found to be In the preceding example, the radiation efficiency is a r respectable 95.4%. However, the real part of the L µf (10.44) impedance of the ESD is much less than the Rloss ≈ 6a πσ characteristic impedance of typical transmission lines where a, µ, and σ are the radius, permeability, and (typically 10s to 100s of ohms). Another problem is conductivity of the wire comprising the antenna, and that the reactance of the ESD is very large. Some f is frequency. form of impedance matching is required to efficiently transfer power from a transmitter or transmission line Reactance. The reactance of the ESD is addressed in to this form of antenna. Failure to do so will result in Section 10.4. A suitable expression for this reactance reflection of a large fraction of the power incident on is (see, e.g., Johnson (1993) in “Additional antenna terminals. A common solution is to insert References” at the end of this section): series inductance to reduce – nominally, cancel – the     negative reactance of the ESD. In the preceding 120 Ω L XA ≈ − ln −1 (10.45) example, about 10 µH would be needed. The πL/λ 2a remaining mismatch in the real-valued components of 2 See Section 9.4 for additional information about the Hertzian impedance is then relatively easy to mitigate using dipole. common “real-to-real” impedance matching

10.7. DIRECTIVITY AND GAIN techniques.

177

10.7 Directivity and Gain

Additional Reading: [m0203]

• R.C. Johnson (Ed.), Antenna Systems Handbook (Ch. 4), McGraw-Hill, 1993.

A transmitting antenna does not radiate power uniformly in all directions. Inevitably more power is radiated in some directions than others. Directivity quantifies this behavior. In this section, we introduce the concept of directivity and the related concepts of maximum directivity and antenna gain. Consider an antenna located at the origin. The power radiated in a single direction (θ, φ) is formally zero. This is because a single direction corresponds to a solid angle of zero, which intercepts an area of zero at any given distance from the antenna. Since the power flowing through any surface having zero area is zero, the power flowing in a single direction is formally zero. Clearly we need a different metric of power in order to develop a sensible description of the spatial distribution of power flow. The appropriate metric is spatial power density; that is, power per unit area, having SI base units of W/m2 . Therefore, directivity is defined in terms of spatial power density in a particular direction, as opposed to power in a particular direction. Specifically, directivity in the direction (θ, φ) is: D(θ, φ) ,

S(r) Save (r)

(10.48)

In this expression, S(r) is the power density at (r, θ, φ); i.e., at a distance r in the direction (θ, φ). Save (r) is the average power density at that distance; that is, S(r) averaged over all possible directions at distance r. Since directivity is a ratio of power densities, it is unitless. Summarizing: Directivity is ratio of power density in a specified direction to the power density averaged over all directions at the same distance from the antenna. Despite Equation 10.48, directivity does not depend on the distance from the antenna. To be specific, directivity is the same at every distance r. Even though the numerator and denominator of Equation 10.48 both vary with r, one finds that the distance dependence always cancels because power density and average power density are both

CHAPTER 10. ANTENNAS

178 proportional to r−2 . This is a key point: Directivity is a convenient way to characterize an antenna because it does not change with distance from the antenna. In general, directivity is a function of direction. However, one is often not concerned about all directions, but rather only the directivity in the direction in which it is maximum. In fact it is quite common to use the term “directivity” informally to refer to the maximum directivity of an antenna. This is usually what is meant when the directivity is indicated to be a single number; in any event, the intended meaning of the term is usually clear from context. Example 10.6. Directivity of the electrically-short dipole. An electrically-short dipole (ESD) consists of a straight wire having length L ≪ λ/2. What is the directivity of the ESD? Solution. The field radiated by an ESD is derived in Section 9.5. In that section, we find that the electric field intensity in the far field of a ˆ-oriented ESD located at the origin is: z −jβr ˆ I0 · βL (sin θ) e e E(r) ≈ θjη 8π r

(10.49)

2

2

|I0 | (βL) 2 1 (sin θ) 2 128π 2 r

(10.50)

and we subsequently find that the total power radiated is: 2

Prad ≈ η

|I0 | (βL) 48π

2

Save =

Prad |I0 | (βL) ≈η 4πr2 192π 2 r2

2

(10.52)

Finally the directivity is determined by applying the definition: D(θ, φ) ,

S(r) Save (r)

≈ 1.5 (sin θ)

(10.53) 2

(10.54)

The maximum directivity occurs in the θ = π/2 plane. Therefore, the maximum directivity is 1.5, meaning the maximum power density is 1.5 times greater than the power density averaged over all directions. Since directivity is a unitless ratio, it is common to express it in decibels. For example, the maximum directivity of the ESD in the preceding example is 10 log10 1.5 ∼ = 1.76 dB. (Note “10 log10 ” here since directivity is the ratio of power-like quantities.) Gain. The gain G(θ, φ) of an antenna is its directivity modified to account for loss within the antenna. Specifically: S(r) for actual antenna Save (r) for identical but lossless antenna (10.55) In this equation, the numerator is the actual power density radiated by the antenna, which is less than the nominal power density due to losses within the antenna. The denominator is the average power density for an antenna which is identical, but lossless. Since the actual antenna radiates less power than an identical but lossless version of the same antenna, gain in any particular direction is always less than directivity in that direction. Therefore, an equivalent definition of antenna gain is G(θ, φ) ,

where I0 represents the magnitude and phase of the current applied to the terminals, η is the wave impedance of the medium, and β = 2π/λ. In Section 10.2, we find that the power density of this field is: S(r) ≈ η

Then:

2

(10.51)

The average power density Save is simply the total power divided by the area of a sphere centered on the ESD. Let us place this sphere at distance r, with r ≫ L and r ≫ λ as required for the validity of Equations 10.49 and 10.50.

G(θ, φ) , erad D(θ, φ)

(10.56)

where erad is the radiation efficiency of the antenna (Section 10.5). Gain is directivity times radiation efficiency; that is, directivity modified to account for loss within the antenna.

10.8. RADIATION PATTERN The receive case. To conclude this section, we make one additional point about directivity, which applies equally to gain. The preceding discussion has presumed an antenna which is radiating; i.e., transmitting. Directivity can also be defined for the receive case, in which it quantifies the effectiveness of the antenna in converting power in an incident wave to power in a load attached to the antenna. Receive directivity is formally introduced in Section 10.13 (“Effective Aperture”). When receive directivity is defined as specified in Section 10.13, it is equal to transmit directivity as defined in this section. Thus, it is commonly said that the directivity of an antenna is the same for receive and transmit. Additional Reading: • “Directivity” on Wikipedia.

179

10.8 Radiation Pattern [m0205]

The radiation pattern of a transmitting antenna describes the magnitude and polarization of the field radiated by the antenna as a function of angle relative to the antenna. A pattern may also be defined for a receiving antenna, however, we defer discussion of the receive case to a later section. The concept of radiation pattern is closely related to the concepts of directivity and gain (Section 10.7). The principal distinction is the explicit consideration of polarization. In many applications, the polarization of the field radiated by a transmit antenna is as important as the power density radiated by the antenna. For example, a radio communication link consists of an antenna which is transmitting separated by some distance from an antenna which is receiving. Directivity determines the power density delivered to the receive antenna, but the receive antenna must be co-polarized with the arriving wave in order to capture all of this power. The radiation pattern concept is perhaps best explained by example. The simplest antenna encountered in common practice is the electrically-short dipole (ESD), which consists of a straight wire of length L that is much less than one-half of a wavelength. In Section 9.5, it is shown that the field radiated by an ESD which is located at the origin and aligned along the z-axis is: −jβr ˆ I0 · βL (sin θ) e e E(r) ≈ θjη 8π r

(10.57)

where I0 represents the magnitude and phase of the current applied to the terminals, η is the wave impedance of the medium, and β = 2π/λ is the phase propagation constant of the medium. Note that the reference direction of the electric field is in the θˆ ˆ direction; therefore, a receiver must be θ-polarized relative to the transmitter’s coordinate system in order to capture all available power. A receiver which is not ˆ fully θ-polarized will capture less power. In the ˆ extreme, a receiver which is φ-polarized with respect to the transmitter’s coordinate system will capture ˆ-oriented ESD, we refer to zero power. Thus, for the z ˆ the θ-polarization of the transmitted field as

180

CHAPTER 10. ANTENNAS

“co-polarized” or simply “co-pol,” and the ˆ φ-polarization of the transmitted field as “cross-pol.” At this point, the reader may wonder what purpose is served by defining cross polarization, since the definition given above seems to suggest that cross-pol should always be zero. In common engineering practice, cross-pol is non-zero when co-pol is different from the intended or nominal polarization of the field radiated by the antenna. In the case of the ESD, we observe that the electric field is always ˆ θ-polarized, and therefore we consider that to be the nominal polarization. Since the actual polarization of the ESD in the example is precisely the same as the nominal polarization of the ESD, the cross-pol of the ideal ESD is zero. If, on the other hand, we arbitrarily define θˆ to be the nominal polarization and apply this definition to a different antenna that does not produce ˆ a uniformly θ-polarized electric field, then we observe non-zero cross-pol. Cross-pol is similarly used to quantify effects due to errors in position or orientation, or due to undesired modification of the field due to materials (e.g., feed or mounting structures) near the transmit antenna. Summarizing: Co-pol is commonly defined to be the intended or nominal polarization for a particular application, which is not necessarily the actual polarization radiated by the antenna under consideration. Cross-pol measures polarization in the orthogonal plane; i.e., deviation from the presumed copol. e Returning to the ESD: Since E(r) depends only on θ and not φ, the co-pol pattern is the same in any plane that contains the z axis. We refer to any such plane as the E-plane. In general:

c S. Lally CC BY-SA 4.0

ˆ-oriented Figure 10.5: E-plane co-pol pattern for the z ESD. In the unnormalized pattern scaling, the dashed circle indicates the maximum value of Equation 10.58. pattern. The associated normalized pattern is addressed later in this section. Similarly, we define the H-plane as follows: The H-plane is any plane in which the nominal or e lies, and so is perpendicular to intended vector H both the E-plane and the direction of propagation. In the case of the ESD, the one and only H-plane is the z = 0 plane. The H-plane pattern is shown in Figure 10.6.

It is often useful to normalize the pattern, meaning that we scale the pattern so that its maximum magnitude corresponds to a convenient value. There The E-plane is any plane in which the nominal or are two common scalings. One common scaling sets e lies. intended vector E the maximum value of the pattern equal to 1, and is therefore independent of source magnitude and distance. This is referred to as the normalized pattern. ˆ Since the ESD is θ-polarized, the E-plane pattern of The normalized co-pol pattern can be defined as the ESD is simply: follows: I0 · βL 1 e e (sin θ) (10.58) E(r) ≈ η ˆ · E(r) e 8π r F (θ, φ) , (10.59) e ˆ · E(r) e This pattern is shown in Figure 10.5. max

Equation 10.58 is referred to as an unnormalized

ˆ is the co-pol reference direction, and the where e

10.8. RADIATION PATTERN

181

c T. Truckle CC BY-SA 4.0 (modified)

c S. Lally CC BY-SA 4.0

ˆ-oriented Figure 10.7: Co-pol pattern typical of a highly direcFigure 10.6: H-plane co-pol pattern for the z ESD. In the unnormalized pattern scaling, the radius tional antenna, such as a yagi. of the circle is the maximum value of Equation 10.58. denominator is the maximum value of the electric field at distance r. A normalized pattern is scaled to a maximum magnitude of 1, using the definition expressed in Equation 10.59.

The other common scaling for patterns sets the maximum value equal to maximum directivity. Directivity is proportional to power density, which is e 2 proportional to E(r) . Therefore, this “directivity normalized” pattern can be expressed as: Dmax |F (θ, φ)|

2

(10.62)

Note that the value of r is irrelevant, since numerator and denominator both scale with r in the same way. Thus, the normalized pattern, like directivity, does not change with distance from the antenna.

where Dmax is the directivity in whichever direction it is maximum. For the ESD considered previously, Dmax = 1.5 (Section 10.7). Therefore, the co-pol pattern of the ESD using this particular scaling is

ˆ so the normalized co-pol pattern ˆ = θ, For the ESD, e is: ˆ e θ · E(r) F (θ, φ) = (ESD) (10.60) e θˆ · E(r)

1.5 sin2 θ (ESD)

max

which yields simply

F (θ, φ) ≈ sin θ (ESD)

(10.63)

Thus, the E-plane co-pol pattern of the ESD using this scaling is similar to Figure 10.5, except that it is squared in magnitude and the radius of the maximum value circle is equal to 1.5, which is 1.76 dB. The H-plane co-pol pattern of the ESD, using this scaling, is Figure 10.6 where the radius of the circle is equal (10.61) to 1.5 (1.76 dB).

Thus, the E-plane normalized co-pol pattern of the ESD is Figure 10.5 where the radius of the maximum value circle is equal to 1, which is 0 dB. Similarly, the H-plane normalized co-pol pattern of the ESD is Figure 10.6 where the radius of the circle is equal to 1 (0 dB).

Pattern lobes. Nearly all antennas in common use exhibit directivity that is greater than that of the ESD. The patterns of these antennas subsequently exhibit more complex structure. Figure 10.7 shows an example. The region around the direction of maximum directivity is referred to as the main lobe.

CHAPTER 10. ANTENNAS

182

the maximum directivity. For the ESD, this plane is the H-plane, so the ESD is said to be omnidirectional in the H-plane. The term “omnidirectional” does not indicate constant pattern in all directions. (In fact, note that the ESD exhibits pattern nulls in two directions.) An omnidirectional antenna, such as the ESD, exhibits constant and nominally maximum directivity in one plane.

c S. Lally CC BY-SA 4.0

Figure 10.8: Half-power beamwidth (HPBW).

The main lobe is bounded on each side by a null, where the magnitude reaches a local minimum, perhaps zero. Many antennas also exhibit a lobe in the opposite direction, known as a backlobe. (Many other antennas exhibit a null in this direction.) Lobes between the main lobe and the backlobe are referred to as sidelobes. Other commonly-used pattern metrics include first sidelobe level, which is the ratio of the maximum magnitude of the sidelobe closest to the main lobe to that of the main lobe; and front-to-back ratio, which is the ratio of the maximum magnitude of the backlobe to that of the main lobe. When the main lobe is narrow, it is common to characterize the pattern in terms of half-power beamwidth (HPBW). This is shown in Figure 10.8. HPBW is the width of the main lobe measured between two points at which the directivity is one-half its maximum value. As a pattern metric, HPBW depends on the plane in which it is measured. In particular, HPBW may be different in the E- and H-planes. For example, the E-plane HPBW of the ESD is found to be 90◦ , whereas the H-plane HPBW is undefined since the pattern is constant in the H-plane. Omnidirectional and isotropic antennas. The ESD is an example of an omnidirectional antenna. An omnidirectional antenna is an antenna whose pattern magnitude is nominally constant in a plane containing

An antenna whose pattern is uniform in all directions is said to be isotropic. No physically-realizable antenna is isotropic; in fact, the ESD is about as close as one can get. Nevertheless, the “isotropic antenna” concept is useful as a standard against which other antennas can be quantified. Since the pattern of an isotropic antenna is constant with direction, the power density radiated by such an antenna in any direction is equal to the power density averaged over all directions. Thus, the directivity of an isotropic antenna is exactly 1. The directivity of any other antenna may be expressed in units of “dBi,” which is simply dB relative to that of an isotropic antenna. For example, the maximum directivity of the ESD is 1.5, which is 10 log10 (1.5/1) = 1.76 dBi. Summarizing: An isotropic antenna exhibits constant directivity in every direction. Such an antenna is not physically-realizable, but is nevertheless useful as a baseline for describing other antennas. Additional Reading: • “Radiation pattern” on Wikipedia.

10.9. EQUIVALENT CIRCUIT MODEL FOR RECEPTION

10.9 Equivalent Circuit Model for Reception [m0206]

In this section, we begin to address antennas as devices that convert incident electromagnetic waves into potentials and currents in a circuit. It is convenient to represent this process in the form of a Th´evenin equivalent circuit. The particular circuit addressed in this section is shown in Figure 10.9. The circuit consists of a voltage source VeOC and a series impedance ZA . The source potential VeOC is the potential at the terminals of the antenna when there is no termination; i.e., when the antenna is open-circuited. The series impedance ZA is the output impedance of the circuit, and so determines the magnitude and phase of the current at the terminals once a load is connected. Given VeOC and the current through the equivalent circuit, it is possible to determine the power delivered to the load. Thus, this model is quite useful, but only if we are able to determine VeOC and ZA . This section provides an informal derivation of these quantities that is sufficient to productively address the subsequent important topics of effective aperture and impedance matching of receive antennas.3 Vector effective length. With no derivation required, we can deduce the following about VeOC : • VeOC must depend on the incident electric field

3 Formal derivations of these quantities are provided in subsequent sections. The starting point is the section on reciprocity.

c S. Lally, CC BY-SA 4.0 (modified)

183

e i . Presumably the relationship is intensity E linear, so VeOC is proportional to the magnitude e i. of E

e i is a vector whereas VeOC is a scalar, • Since E there must be some vector le for which e i · le VeOC = E

(10.64)

e i has SI base units of V/m and VeOC has • Since E SI base units of V, le must have SI base units of m; i.e., length. • We expect that VeOC increases as the size of the antenna increases, so the magnitude of le likely increases with the size of the antenna. • The direction of le must be related to the orientation of the incident electric field relative to that of the antenna, since this is clearly important yet we have not already accounted for this. It may seem at this point that le is unambiguously determined, and we need merely to derive its value. However, this is not the case. There are in fact multiple unique definitions of le that will reduce the e i to the observed scalar VeOC via vector E Equation 10.64. In this section, we shall employ the most commonly-used definition, in which le is referred to as vector effective length. Following this definition, the scalar part le of le = ˆlle is commonly referred to as any of the following: effective length (the term used in this book), effective height, or antenna factor. In this section, we shall merely define vector effective length, and defer a formal derivation to Section 10.11. In this definition, we arbitrarily set ˆl, the real-valued unit vector indicating the direction of le , equal to the direction in which the electric field transmitted from this antenna would be polarized in the far field. For ˆ-oriented electrically-short example, consider a z dipole (ESD) located at the origin. The electric field transmitted from this antenna would have only a θˆ component, and no φˆ component (and certainly no ˆr component). Thus, ˆl = θˆ in this case.

e i · ˆl yields the scalar Figure 10.9: Th´evenin equivalent circuit for an an- Applying this definition, E i e tenna in the presence of an incident electromagnetic component of E that is co-polarized with electric wave. field radiated by the antenna when transmitting. Now

CHAPTER 10. ANTENNAS

184 le is uniquely defined to be the factor that converts this component into VeOC . Summarizing:

The vector effective length le = ˆlle is defined as follows: ˆl is the real-valued unit vector corresponding to the polarization of the electric field that would be transmitted from the antenna in the far field. Subsequently, the effective length le is le ,

VeOC e i · ˆl E

(10.65)

where VeOC is the open-circuit potential induced at the antenna terminals in response to the incie i. dent electric field intensity E While this definition yields an unambiguous value for le , it is not yet clear what that value is. For most antennas, effective length is quite difficult to determine directly, and one must instead determine effective length indirectly from the transmit characteristics via reciprocity. This approach is relatively easy (although still quite a bit of effort) for thin dipoles, and is presented in Section 10.11. To provide an example of how effective length works ˆ-oriented ESD described right away, consider the z earlier in this section. Let the length of this ESD be ˆ e i be a θ-polarized L. Let E plane wave arriving at the ESD. The ESD is open-circuited, so the potential induced in its terminals is VeOC . One observes the following: e i arrives from anywhere in the θ = π/2 • When E e i points in plane (i.e., broadside to the ESD), E the −ˆ z direction, and we find that le ≈ L/2. It should not be surprising that le is proportional to L; this expectation was noted earlier in this section. e i arrives from the directions θ = 0 or • When E e i is θ = π – i.e., along the axis of the ESD – E perpendicular to the axis of the ESD. In this case, we find that le equals zero. Taken together, these findings suggest that le should contain a factor of sin θ. We conclude that the vector

ˆ-directed ESD of length L is effective length for a z L le ≈ θˆ sin θ 2

(ESD)

(10.66)

Example 10.7. Potential induced in an ESD. A thin straight dipole of length 10 cm is located at the origin and aligned with the z-axis. A plane wave is incident on the dipole from the direction (θ = π/4, φ = π/2). The frequency of the wave is 30 MHz. The magnitude of the incident electric field is 10 µV/m (rms). What is the magnitude of the induced open-circuit potential ˆ when the electric field is (a) θ-polarized and (b) ˆ φ-polarized? Solution. The wavelength in this example is c/f ∼ = 10 m, so this dipole is electrically-short. Using Equation 10.66: 10 cm π sin le ≈ θˆ 2 4 ˆ ≈ θ (3.54 cm) Thus, the effective length le = 3.54 cm. When ˆ the electric field is θ-polarized, the magnitude of the induced open-circuit voltage is e ei VOC = E · le ≈ (10 µV/m) θˆ · θˆ (3.54 cm) ≈ 354 nV rms (a)

ˆ When the electric field is φ-polarized: e VOC ≈ (10 µV/m) φˆ · θˆ (3.54 cm) ≈ 0 (b)

This is because the polarization of the incident electric field is orthogonal to that of the ESD. In fact, the answer to part (b) is zero for any angle of incidence (θ, φ). Output impedance. The output impedance ZA is somewhat more difficult to determine without a formal derivation, which is presented in Section 10.12. For the purposes of this section, it suffices to jump directly to the result:

10.9. EQUIVALENT CIRCUIT MODEL FOR RECEPTION The output impedance ZA of the equivalent circuit for an antenna in the receive case is equal to the input impedance of the same antenna in the transmit case. This remarkable fact is a consequence of the reciprocity property of antenna systems, and greatly simplifies the analysis of receive antennas. Now a demonstration of how the antenna equivalent circuit can be used to determine the power delivered by an antenna to an attached electrical circuit: Example 10.8. Power captured by an ESD. Continuing with part (a) of Example 10.7: If this antenna is terminated into a conjugate-matched load, then what is the power delivered to that load? Assume the antenna is lossless. Solution. First, we determine the impedance ZA of the equivalent circuit of the antenna. This is equal to the input impedance of the antenna in transmission. Let RA and XA be the real and imaginary parts of this impedance; i.e., ZA = RA + jXA . Further, RA is the sum of the radiation resistance Rrad and the loss resistance. The loss resistance is zero because the antenna is lossless. Since this is an ESD:  2 L 2 (10.67) Rrad ≈ 20π λ Therefore, RA = Rrad ≈ 4.93 mΩ. We do not need to calculate XA , as will become apparent in the next step. ∗ A conjugate-matched load has impedance ZA , so the potential VeL across the load is

VeL = VeOC

∗ ∗ ZA eOC ZA = V ∗ ZA + ZA 2RA

The current IeL through the load is IeL =

VeOC VeOC = ∗ ZA + ZA 2RA

(10.68)

(10.69)

Taking VeOC as an RMS quantity, the power PL

185

delivered to the load is PL = Re {VL IL∗ } =

e 2 VOC

(10.70) 4RA In part (a) of Example 10.7, VeOC is found to be ≈ 354 nV rms, so PL ≈ 6.33 pW. Additional Reading: • “Th´evenin’s Theorem” on Wikipedia. • W.L. Stutzman & G.A. Thiele, Antenna Theory and Design, 3rd Ed., Wiley, 2012. Sec. 4.2 (“Receiving Properties of Antennas”).

CHAPTER 10. ANTENNAS

186

10.10 Reciprocity [m0214]

The term “reciprocity” refers to a class of theorems that relate the inputs and outputs of a linear system to those of an identical system in which the inputs and outputs are swapped. The importance of reciprocity in electromagnetics is that it simplifies problems that would otherwise be relatively difficult to solve. An example of this is the derivation of the receiving properties of antennas, which is addressed in other sections using results derived in this section. As an initial and relatively gentle introduction, consider a well-known special case that emerges in basic circuit theory: The two-port device shown in Figure 10.10. The two-port is said to be reciprocal if the voltage v2 appearing at port 2 due to a current applied at port 1 is the same as v1 when the same current is applied instead at port 2. This is normally the case when the two-port consists exclusively of linear passive devices such as ideal resistors, capacitors, and inductors. The fundamental underlying requirement is linearity: That is, outputs must be proportional to inputs, and superposition must apply. This result from basic circuit theory is actually a special case of a more general theorem of electromagnetics, which we shall now derive. Figure 10.11 shows a scenario in which a current e1 is completely contained within a distribution J volume V. This current is expressed as a volume current density, having SI base units of A/m2 . Also, the current is expressed in phasor form, signaling that we are considering a single frequency. This current e 1, distribution gives rise to an electric field intensity E

c C. Wang CC BY-SA 4.0

Figure 10.11: An electromagnetic system consisting e1 radiating an electric field of a current distribution J e E1 in the presence of linear matter.

c C. Wang CC BY-SA 4.0

Figure 10.12: The same electromagnetic system as shown in Figure 10.11 (including the same distribution e2 , resulting of matter), but now with a different input, J e in a different output, E2 . Inductiveload, public domain (modified)

Figure 10.10: Two-port device.

10.10. RECIPROCITY having SI base units of V/m. The volume may consist of any combination of linear time-invariant matter; i.e., permittivity ǫ, permeability µ, and conductivity σ are constants that may vary arbitrarily with position but not with time.

187 of the system. However, the answer turns out to be that a surprising bit more information is available. This information is provided by reciprocity. To show this, we must engage in some pure mathematics.

Derivation of the Lorentz reciprocity theorem. e 2 with each side Here’s a key idea: We may interpret this scenario as a First, let us take the dot product of H e1 is the input, E e 1 is the of Equation 10.71: “two-port” system in which J output, and the system’s behavior is completely   e2 · ∇×E e 1 = −jωµH e1 ·H e2 defined by Maxwell’s equations in differential phasor H (10.75) form: e 1 = −jωµH e1 e 1 with each ∇×E (10.71) Similarly, let us take the dot product of E e1 = J e1 + jωǫE e1 ∇×H (10.72) side of Equation 10.74:   e1 · ∇ × H e2 = E e1 · J e2 + jωǫE e1 · E e 2 (10.76) along with the appropriate electromagnetic boundary E conditions. (For the purposes of our present analysis, e 1 is neither an input nor an output; it is merely a H Next, let us subtract Equation 10.75 from coupled quantity that appears in this particular Equation 10.76. The left side of the resulting equation e 1 and J e1 .) formulation of the relationship between E is     e1 · ∇ × H e2 −H e2 · ∇×E e1 Figure 10.12 shows a scenario in which a different E (10.77) e2 is completely contained within current distribution J This can be simplified using the vector identity the same volume V containing the same distribution (Appendix B.3): of linear matter. This new current distribution gives e 2 . The relationship rise to an electric field intensity E ∇ · (A × B) = B · (∇ × A) − A · (∇ × B) (10.78) e 2 and J e2 is governed by the same between E equations: Yielding e 2 = −jωµH e2 ∇×E (10.73)       e1 · ∇ × H e2 −H e2· ∇×E e1 = ∇ · H e2 ×E e1 e2 = J e2 + jωǫE e2 ∇×H (10.74) E (10.79) along with the same electromagnetic boundary So, by subtracting Equation 10.75 from conditions. Thus, we may interpret this scenario as Equation 10.76, we have found: the same electromagnetic system depicted in   e2 as the input and E e2 Figure 10.11, except now with J e2 ×E e1 = E e1 ·J e2 + jωǫE e1 ·E e 2 + jωµH e 1 ·H e2 ∇ · H is the output. (10.80) The input current and output field in the second scenario are, in general, completely different from the Next, we repeat the process represented by Equations 10.75–10.80 above to generate a input current and output field in the first scenario. complementary equation to Equation 10.80. This time However, both scenarios involve precisely the same e 2 with each side of we take the dot product of E electromagnetic system; i.e., the same governing Equation 10.72: equations and the same distribution of matter. This   leads to the following question: Let’s say you know e e e e e e E · ∇ × H (10.81) 2 1 = E2 · J1 + jωǫE1 · E2 nothing about the system, aside from the fact that it is linear and time-invariant. However, you are able to e 1 in e 1 with each observe the first scenario; i.e., you know E Similarly, let us take the dot product of H e response to J1 . Given this very limited look into the side of Equation 10.73: e2 behavior of the system, what can you infer about E e2 , or vice-versa? At first glance, the answer   given J e1 · ∇×E e 2 = −jωµH e1 ·H e2 H (10.82) might seem to be nothing, since you lack a description

CHAPTER 10. ANTENNAS

188 Next, let us subtract Equation 10.82 from Equation 10.81. Again using the vector identity, the left side of the resulting equation is       e2 · ∇ × H e1 −H e1· ∇×E e2 = ∇ · H e1 ×E e2 E (10.83) So we find:   e1 ×E e2 = E e2 ·J e1 + jωǫE e1 ·E e 2 + jωµH e 1 ·H e2 ∇· H (10.84) Finally, subtracting Equation 10.84 from Equation 10.80, we obtain:   e2 ×E e1 − H e1 ×E e2 = E e1 · J e2 − E e2 · J e1 ∇· H

(10.85) This equation is commonly known by the name of the theorem it represents: The Lorentz reciprocity theorem. The theorem is a very general statement about the relationship between fields and currents at each point in space. An associated form of the theorem applies to contiguous regions of space. To obtain this form, we simply integrate both sides of Equation 10.85 over the volume V: Z   e2 ×E e1 − H e1 ×E e 2 dv ∇· H V Z   e1 · J e2 − E e2 · J e1 dv = E (10.86)

non-zero source current and S is infinite. Because field magnitude diminishes with distance from the e 1, H e 1 ) and (E e 2, H e 2 ) are all source, the fields (E effectively zero on S. In this case, the left side of Equation 10.87 is zero, and we find: Z  V

 e1 · J e2 − E e2 · J e1 dv = 0 E

(10.88)

for any volume V which contains all the current. The Lorentz reciprocity theorem (Equation 10.88) describes a relationship between one distribution of current and the resulting fields, and a second distribution of current and resulting fields, when both scenarios take place in identical regions of space filled with identical distributions of linear matter. Why do we refer to this relationship as reciprocity? Simply because the expression is identical when the subscripts “1” and “2” are swapped. In other words, the relationship does not recognize a distinction between “inputs” and “outputs;” there are only “ports.”

A useful special case pertains to scenarios in which e1 and J e2 are spatially the current distributions J disjoint. By “spatially disjoint,” we mean that there is V e1 and J e2 are no point in space at which both J We now take the additional step of using the non-zero; in other words, these distributions do not divergence theorem (Appendix B.3) to transform the overlap. (Note that the currents shown in left side of the equation into a surface integral: Figures 10.11 and 10.12 are depicted as spatially I  disjoint.) To see what happens in this case, let us first  e2 ×E e1 − H e1 ×E e 2 · ds H rewrite Equation 10.88 as follows: S Z  Z Z  e e e e e e e2 · J e1 dv = E1 · J2 − E2 · J1 dv (10.87) E1 · J2 dv = E (10.89) V

where S is the closed mathematical surface which bounds V. This is also the Lorentz reciprocity theorem, but now in integral form. This version of the theorem relates fields on the bounding surface to sources within the volume.

The integral form of the theorem has a particularly useful feature. Let us confine the sources to a finite region of space, while allowing V to grow infinitely large, expanding to include all space. In this situation, the closest distance between any point containing

V

V

e1 is Let V1 be the contiguous volume over which J non-zero, and let V2 be the contiguous volume over e2 is non-zero. Then Equation 10.89 may be which J written as follows: Z Z e e e2 · J e1 dv E1 · J2 dv = E (10.90) V2

V1

The utility of this form is that we have reduced the region of integration to just those regions where the current exists.

10.10. RECIPROCITY

189

c C. Wang CC BY-SA 4.0

Figure 10.13: antennas.

c C. Wang CC BY-SA 4.0

A two-port consisting of two dipole

Figure 10.14: The port 1 antenna (a) transmitting and (b) receiving.

Reciprocity of two-ports consisting of antennas. Equation 10.90 allows us to establish the reciprocity of two-ports consisting of pairs of antennas. This is most easily demonstrated for pairs of thin dipole antennas, as shown in Figure 10.13. Here, port 1 is defined by the terminal quantities (Ve1 , Ie1 ) of the antenna on the left, and port 2 is defined by the terminal quantities (Ve2 , Ie2 ) of the antenna on the right. These quantities are defined with respect to a small gap of length ∆l between the perfectly-conducting arms of the dipole. Either antenna may transmit or receive, so (Ve1 , Ie1 ) depends on (Ve2 , Ie2 ), and vice-versa.

The transmit and receive cases for port 1 are illustrated in Figure 10.14(a) and (b), respectively. e1 is the current distribution on this antenna Note that J when transmitting (i.e., when driven by the impressed e 2 is the electric field incident on the current Ie1t ) and E antenna when the other antenna is transmitting. Let us select V1 to be the cylindrical volume defined by the exterior surface of the dipole, including the gap between the arms of the dipole. We are now ready to consider the right side of Equation 10.90: Z e2 · J e1 dv E (10.91) V1

The electromagnetic boundary conditions applicable to the perfectly-conducting dipole arms allow this integral to be dramatically simplified. First, recall that all current associated with a perfect conductor must flow on the surface of the material, and therefore

e1 = 0 everywhere except on the surface. Therefore, J e1 is always tangent to the surface. the direction of J e 2 is zero on the surface The tangent component of E of a perfectly-conducting material, as required by the applicable boundary condition. Therefore, e2 · J e1 = 0 everywhere J e1 is non-zero. E

There is only one location where the current is non-zero and yet there is no conductor: This location is the gap located precisely at the terminals. Thus, we find: Z Z e2 · J e1 dv = e 2 · Ietˆl1 dl (10.92) E E 1 V1

gap

where the right side is a line integral crossing the gap that defines the terminals of the antenna. We assume the current Ie1t is constant over the gap, and so may be factored out of the integral: Z Z t e e e 2 · ˆl1 dl e E2 · J1 dv = I1 E (10.93) V1

gap

Recall that potential between two points in space is given by the integral of the electric field over any path between those two points. In particular, we may calculate the open-circuit potential Ve1r as follows: Z r e 2 · ˆl1 dl e V1 = − E (10.94) gap

Remarkably, we have found: Z e2 · J e1 dv = −Iet Ve r E 1 1 V1

(10.95)

CHAPTER 10. ANTENNAS

190 Applying the exact same procedure for port 2 (or by simply exchanging subscripts), we find: Z e1 · J e2 dv = −Iet Ve r E (10.96) 2 2 V2

10.11 Potential Induced in a Dipole [m0215]

Now substituting these results into Equation 10.90, An electromagnetic wave incident on an antenna will we find: induce a potential at the terminals of the antenna. In Ie1t Ve1r = Ie2t Ve2r (10.97) this section, we shall derive this potential. To simplify the derivation, we shall consider the special case of a At the beginning of this section, we stated that a e two-port is reciprocal if V2 appearing at port 2 due to straight thin dipole of arbitrary length that is a current applied at port 1 is the same as Ve1 when the illuminated by a plane wave. However, certain aspects of the derivation will apply to antennas generally. In same current is applied instead at port 2. If the r e present two-dipole problem is reciprocal, then V2 due particular, the concepts of effective length (also known as effective height) and vector effective length to Ie1t should be the same as Ve1r when Ie2t = Ie1t . Is it? emerge naturally from this derivation, so this section t Let us set Ie1 equal to some particular value I0 , then also serves as a stepping stone in the development of the resulting value of Ve2r will be some particular an equivalent circuit model for a receiving antenna. value V0 . If we subsequently set Ie2t equal to I0 , then The derivation relies on the transmit properties of the resulting value of Ve1r will be, according to dipoles as well as the principle of reciprocity, so Equation 10.97: familiarity with those topics is recommended before reading this section. I 0 V0 Ie2t Ve2r r e = V0 (10.98) = V1 = The scenario of interest is shown in Figure 10.15. I0 Ie1t ˆ-aligned straight dipole is located at the Here a thin z Therefore, Equation 10.97 is simply a mathematical origin. The total length of the dipole is L. The arms form of the familiar definition of reciprocity from of the dipole are perfectly-conducting. The terminals basic circuit theory, and we have found that our consist of a small gap of length ∆l between the arms. system of antennas is reciprocal in precisely this same The incident plane wave is described in terms of its e i . The question is: What is sense. electric field intensity E e VOC , the potential at the terminals when the terminals The above analysis presumed pairs of straight, are open-circuited? perfectly conducting dipoles of arbitrary length. However, the result is readily generalized – in fact, is There are multiple approaches to solve this problem. the same – for any pair of passive antennas in linear A direct attack is to invoke the principle that potential time-invariant media. Summarizing: is equal to the integral of the electric field intensity over a path. In this case, the path begins at the “−” The potential induced at the terminals of one anterminal and ends at the “+” terminal, crossing the tenna due to a current applied to a second antenna gap that defines the antenna terminals. Thus: is equal to the potential induced in the second anZ tenna by the same current applied to the first ane gap · dl VeOC = − E (10.99) tenna (Equation 10.97). gap Additional Reading: • “Reciprocity (electromagnetism)” on Wikipedia. • “Two-port network” on Wikipedia.

e gap is the electric field in the gap. The where E e gap problem with this approach is that the value of E e i , because is not readily available. It is not simply E the antenna structure (in particular, the electromagnetic boundary conditions) modify the electric field in the vicinity of the antenna.4 4 Also,

if this were true, then the antenna itself would not matter;

10.11. POTENTIAL INDUCED IN A DIPOLE

191

Fortunately, we can bypass this obstacle using the principle of reciprocity. In a reciprocity-based strategy, we establish a relationship between two scenarios that take place within the same electromagnetic system. The first scenario is shown in Figure 10.16. In this scenario, we have two dipoles. The first dipole is precisely the dipole of interest (Figure 10.15), except that a current Ie1t is applied to the antenna terminals. This gives rise to a current e (SI base units of A) along the dipole, distribution I(z) and subsequently the dipole radiates the electric field (Section 9.6): −jβr

ˆ ηe e 1 (r) ≈ θj E (sin θ) # " 2Z r 1 +L/2 e ′ +jβz′ cos θ ′ I(z )e dz (10.100) · λ −L/2

ˆ The second antenna is a θ-aligned Hertzian dipole in e 1 . (For a refresher on the far field, which receives E the properties of Hertzian dipoles, see Section 9.4. A key point is that Hertzian dipoles are vanishingly small.) Specifically, we measure (conceptually, at least) the open-circuit potential Ve2r at the terminals of the Hertzian dipole. We select a Hertzian dipole for this purpose because – in contrast to essentially all

c C. Wang CC BY-SA 4.0

Figure 10.16: The dipole of interest driven by current e 1 , resulting in open-circuit Ie1t radiates electric field E r e potential V2 at the terminals of a Hertzian dipole in the far field.

only the relative spacing and orientation of the antenna terminals would matter!

c C. Wang CC BY-SA 4.0

Figure 10.17: The Hertzian dipole driven by current e 2 , resulting in open-circuit Ie2t radiates electric field E Figure 10.15: A potential is induced at the terminals potential Ve r at the terminals of the dipole of interest. 1 of a thin straight dipole in response to an incident plane wave. c C. Wang CC BY-SA 4.0

CHAPTER 10. ANTENNAS

192 other antennas – it is simple to determine the open circuit potential. As explained earlier: Z e gap · dl Ve2r = − E (10.101) gap

Now we ask: What is the induced potential Ve1r in the dipole of interest? Once again, Equation 10.99 is not much help, because the electric field in the gap is not known. However, we do know that Ve1r should be e 2 (r = 0), since this is presumed to proportional to E be a linear system. Based on this much information alone, there must be some vector le = ˆlle for which

e gap is simply the incident For the Hertzian dipole, E electric field, since there is negligible structure (in particular, a negligible amount of material) present to e 2 (r = 0) · le Ve1r = E (10.106) modify the electric field. Thus, we have simply: Z This does not uniquely define either the unit vector ˆl r e e nor the scalar part le , since a change in the definition V2 = − E1 · dl (10.102) gap of the former can be compensated by a change in the definition of the latter and vice-versa. So at this point Since the Hertzian dipole is very short and very far we invoke the standard definition of le as the vector e 1 is essentially away from the transmitting dipole, E effective length, introduced in Section 10.9. Thus, ˆl is constant over the gap. Also recall that we required the defined to be the direction in which an electric field e 1 . Choosing to Hertzian dipole to be aligned with E transmitted from the antenna would be polarized. In integrate in a straight line across the gap, ˆl = −θ, ˆ where the minus sign the present example, Equation 10.102 reduces to: reflects the fact that positive terminal potential results ˆ e 1 (r2 ) · θ∆l z direction. (10.103) in terminal current which flows in the −ˆ Ve2r = −E Thus, Equation 10.106 becomes: where ∆l is the length of the gap and r2 is the ˆe e 2 (r = 0) · θl Ve1r = −E (10.107) location of the Hertzian dipole. Substituting the e 1 from Equation 10.100, we obtain: expression for E We may go a bit further and substitute the expression −jβr2 e 2 (r = 0) from Equation 10.105: for E e η (sin θ) Ve2r ≈ − j 2 r " Z2 # Iet · β∆l e−jβr2 le (10.108) Ve1r ≈ −jη 2 1 +L/2 e ′ +jβz′ cos θ ′ 4π r2 · I(z )e dz ∆l λ −L/2 where r2 = |r2 |.

(10.104) Now we invoke reciprocity. As a two-port linear time-invariant system, it must be true that:

The second scenario is shown in Figure 10.17. This scenario is identical to the first scenario, with the exception that the Hertzian dipole transmits and the dipole of interest receives. The field radiated by the Hertzian dipole in response to applied current Ie2t , evaluated at the origin, is (Section 9.4): −jβr2 et ˆ I2 · β∆l (1) e e 2 (r = 0) ≈ θjη E 4π r2

(10.105)

Thus:

Ie1t Ve1r = Ie2t Ve2r Iet Ve1r = 2 Ve2r Ie1t

(10.109)

(10.110)

Substituting the expression for Ve2r from Equation 10.104:

Iet η e−jβr2 Ve1r ≈ − 2 · j (sin θ) 2 r2 Ie1t " Z # 1 +L/2 e ′ +jβz′ cos θ ′ · I(z )e dz ∆l λ −L/2

The “sin θ” factor in the general expression is equal to 1 in this case, since, as shown in Figure 10.17, the origin is located broadside (i.e., at π/2 rad) relative to the axis of the Hertzian dipole. Also note that because (10.111) the Hertzian dipole is presumed to be in the far field, E2 may be interpreted as a plane wave in the region Thus, reciprocity has provided a second expression of the receiving dipole of interest. for Ve1r . We may solve for le by setting this expression

10.11. POTENTIAL INDUCED IN A DIPOLE equal to the expression from Equation 10.108, yielding " Z # 2π 1 +L/2 e ′ +jβz′ cos θ ′ le ≈ I(z )e dz sin θ β Ie1t λ −L/2 (10.112) Noting that β = 2π/λ, this simplifies to: "

1 le ≈ e It

1

Z

+L/2 −L/2

# ′ +jβz ′ cos θ ′ e I(z )e dz sin θ

(10.113) Thus, you can calculate le using the following procedure: 1. Apply a current Ie1t to the dipole of interest.

e 2. Determine the resulting current distribution I(z) along the length of the dipole. (Note that precisely this is done for the electrically-short dipole in Section 9.5 and for the half-wave dipole in Section 9.7.) e over the length of the dipole as 3. Integrate I(z) indicated in Equation 10.113. Then divide e (“normalize”) by Ie1t (which is simply I(0)). Note that the result is independent of the excitation Ie1t , as expected since this is a linear system. 4. Multiply by sin θ. We have now determined that the open-circuit terminal potential VeOC in response to an incident e i is electric field E e i · le VeOC = E

(10.114)

where le = ˆlle is the vector effective length defined previously. This result is remarkable. In plain English, we have found that: The potential induced in a dipole is the copolarized component of the incident electric field times a normalized integral of the transmit current distribution over the length of the dipole, times sine of the angle between the dipole axis and the direction of incidence.

193 In other words, the reciprocity property of linear systems allows this property of a receiving antenna to be determined relatively easily if the transmit characteristics of the antenna are known. Example 10.9. Effective length of a thin electrically-short dipole (ESD). A explained in Section 9.5, the current distribution on a thin ESD is   e ≈ I0 1 − 2 |z| (10.115) I(z) L

where L is the length of the dipole and I0 is the terminal current. Applying Equation 10.113, we find: " Z #   1 +L/2 2 ′ +jβz ′ cos θ ′ dz le ≈ I0 1 − |z | e I0 −L/2 L · sin θ

(10.116)

Recall β = 2π/λ, so βz ′ = 2π (z ′ /λ). Since this is an electrically-short dipole, z ′ ≪ λ over the entire integral, and subsequently we may ′ assume e+jβz cos θ ≈ 1 over the entire integral. Thus: "Z  # +L/2  2 ′ le ≈ 1 − |z | dz ′ sin θ L −L/2 (10.117) The integral is easily solved using standard methods, or simply recognize that the “area under the curve” in this case is simply one-half “base” (L) times “height” (1). Either way, we find L le ≈ sin θ (10.118) 2 Example 10.7 (Section 10.9) demonstrates how Equation 10.114 with the vector effective length determined in the preceding example is used to obtain the induced potential.

CHAPTER 10. ANTENNAS

194

10.12 Equivalent Circuit Model for Reception, Redux [m0216]

Section 10.9 provides an informal derivation of an equivalent circuit model for a receiving antenna. This model is shown in Figure 10.18. The derivation of this model was informal and incomplete because the open-circuit potential Ei · le and source impedance ZA were not rigorously derived in that section. While the open-circuit potential was derived in Section 10.11 (“Potential Induced in a Dipole”), the source impedance has not yet been addressed. In this section, the source impedance is derived, which completes the formal derivation of the model. Before reading this section, a review of Section 10.10 (“Reciprocity”) is recommended. The starting point for a formal derivation is the two-port model shown in Figure 10.19. If the two-port is passive, linear, and time-invariant, then the potential v2 is a linear function of potentials and currents present at ports 1 and 2. Furthermore, v1 must be proportional to i1 , and similarly v2 must be proportional to i2 , so any pair of “inputs” consisting of either potentials or currents completely determines the two remaining potentials or currents. Thus, we may write:

Inductiveload, public domain (modified)

Figure 10.19: Two-port system. these quantities as impedances. Similarly we may write: v1 = Z11 i1 + Z12 i2 (10.120) We can develop expressions for Z12 and Z21 as follows. First, note that v1 = Z12 i2 when i1 = 0. Therefore, we may define Z12 as follows: v1 Z12 , (10.121) i2 i1 =0

The simplest way to make i1 = 0 (leaving i2 as the sole “input”) is to leave port 1 open-circuited. In previous sections, we invoked special notation for these circumstances. In particular, we defined Ie2t as i2 in phasor representation, with the superscript “t” (“transmitter”) indicating that this is the sole “input”; and Ve1r as v1 in phasor representation, with the superscript “r” (“receiver”) signaling that port 1 is v2 = Z21 i1 + Z22 i2 (10.119) both open-circuited and the “output.” Applying this notation, we note: where Z11 and Z12 are, for the moment, simply constants of proportionality. However, note that Z11 Ve r (10.122) Z12 = 1 and Z12 have SI base units of Ω, and so we refer to Ie2t Similarly:

Z21 =

Ve2r Iet

(10.123)

1

In Section 10.10 (“Reciprocity”), we established that a pair of antennas could be represented as a passive linear time-invariant two-port, with v1 and i1 representing the potential and current at the terminals c S. Lally CC BY-SA 4.0

of one antenna (“antenna 1”), and v2 and i2 Figure 10.18: Th´evenin equivalent circuit model for representing the potential and current at the terminals an antenna in the presence of an incident electric field of another antenna (“antenna 2”). Therefore for any Ei . pair of antennas, quantities Z11 , Z12 , Z22 , and Z21

10.12. EQUIVALENT CIRCUIT MODEL FOR RECEPTION, REDUX can be identified that completely determine the relationship between the port potentials and currents.

Therefore: i2 = −

We also established in Section 10.10 that:

Therefore,

Ie1t Ve1r = Ie2t Ve2r

(10.124)

Ve1r Ve r = 2 (10.125) Ie2t Ie1t Referring to Equations 10.122 and 10.123, we see that Equation 10.125 requires that: Z12 = Z21

(10.126)

This is a key point. Even though we derived this equality by open-circuiting ports one at a time, the equality must hold generally since Equations 10.119 and 10.120 must apply – with the same values of Z12 and Z21 – regardless of the particular values of the port potentials and currents. We are now ready to determine the Th´evenin equivalent circuit for a receiving antenna. Let port 1 correspond to the transmitting antenna; that is, i1 is Ie1t . Let port 2 correspond to an open-circuited receiving antenna; thus, i2 = 0 and v2 is Ve2r . Now applying Equation 10.119: v2 = Z21 i1 + Z22 i2   = Ve2r /Ie1t Ie1t + Z22 · 0 = Ve2r

(10.127)

We previously determined Ve2r from electromagnetic considerations to be (Section 10.11): e i · le Ve2r = E

(10.128)

e i is the field incident on the receiving where E antenna, and le is the vector effective length as defined in Section 10.11. Thus, the voltage source in the Th´evenin equivalent circuit for the receive e i · le , as shown in Figure 10.18. antenna is simply E

195

Z21 et I Z22 1

(10.130)

Now using Equation 10.123 to eliminate Ie1t , we obtain: Ve r i2 = − 2 (10.131) Z22 Note that the reference direction for i2 as defined in Figure 10.19 is opposite the reference direction for short-circuit current. That is, given the polarity of v2 shown in Figure 10.19, the reference direction of current flow through a passive load attached to this port is from “+” to “−” through the load. Therefore, the source impedance, calculated as the ratio of the open-circuit potential to the short circuit current, is: Ve2r = Z22 +Ve r /Z22

(10.132)

2

We have found that the series impedance ZA in the Th´evenin equivalent circuit is equal to Z22 in the two-port model. To determine Z22 , let us apply a current i2 = Ie2t to port 2 (i.e., antenna 2). Equation 10.119 indicates that we should see: v2 = Z21 i1 + Z22 Ie2t

(10.133)

Solving for Z22 :

Z22 =

v2 i1 − Z21 t e I Iet 2

(10.134)

2

Note that the first term on the right is precisely the impedance of antenna 2 in transmission. The second term in Equation 10.134 describes a contribution to Z22 from antenna 1. However, our immediate interest is in the equivalent circuit for reception of an electric The other component in the Th´evenin equivalent e i in the absence of any other antenna. We can field E circuit is the series impedance. From basic circuit e i is generated have it both ways by imagining that E theory, this impedance is the ratio of v2 when port 2 is by antenna 1, but also that antenna 1 is far enough open-circuited (i.e., Ve2r ) to i2 when port 2 is away to make Z21 – the factor that determines the short-circuited. This value of i2 can be obtained using effect of antenna 1 on antenna 2 – negligible. Then Equation 10.119 with v2 = 0: we see from Equation 10.134 that Z22 is the 0 = Z21 Ie1t + Z22 i2 (10.129) impedance of antenna 2 when transmitting.

CHAPTER 10. ANTENNAS

196 Summarizing: The Th´evenin equivalent circuit for an antenna e i is in the presence of an incident electric field E shown in Figure 10.18. The series impedance ZA in this model is equal to the impedance of the antenna in transmission. Mutual coupling. This concludes the derivation, but raises a follow-up question: What if antenna 2 is present and not sufficiently far away that Z21 can be assumed to be negligible? In this case, we refer to antenna 1 and antenna 2 as being “coupled,” and refer to the effect of the presence of antenna 1 on antenna 2 as coupling. Often, this issue is referred to as mutual coupling, since the coupling affects both antennas in a reciprocal fashion. It is rare for coupling to be significant between antennas on opposite ends of a radio link. This is apparent from common experience. For example, changes to a receive antenna are not normally seen to affect the electric field incident at other locations. However, coupling becomes important when the antenna system is a dense array; i.e., multiple antennas separated by distances less than a few wavelengths. It is common for coupling among the antennas in a dense array to be significant. Such arrays can be analyzed using a generalized version of the theory presented in this section. Additional Reading: • “Th´evenin’s Theorem” on Wikipedia.

10.13 Effective Aperture [m0218]

When working with systems involving receiving antennas, it is convenient to have a single parameter that relates incident power (as opposed to incident electric field) to the power delivered to the receiver electronics. This parameter is commonly known as the effective aperture or antenna aperture. From elementary circuit theory, the power delivered to a load ZL is maximized when the load is conjugate matched to the antenna impedance; i.e., when ∗ ZL = ZA where ZA is the impedance of the antenna. Thus, a convenient definition for effective aperture Ae employs the relationship: i PR,max , Sco Ae

(10.135)

i where Sco is the incident power density (SI base units 2 of W/m ) that is co-polarized with the antenna, and PR,max is the power delivered to a load that is conjugate matched to the antenna impedance. Summarizing:

Effective aperture (SI base units of m2 ) is the ratio of power delivered by an antenna to a conjugate matched load, to the incident co-polarized power density. As defined, a method for calculation of effective aperture is not obvious, except perhaps through direct measurement. In fact, there are at least three ways we can calculate effective aperture: (a) via effective length, (b) via thermodynamics, and (c) via reciprocity. Each of these methods yields some insight and are considered in turn. Effective aperture via effective length. The potential and current at the load of the receive antenna can be determined using the equivalent circuit model shown in Figure 10.20, using the Th´evenin equivalent circuit model for the antenna developed in Section 10.9. In this model, the voltage source is ei determined by the incident electric field intensity E ˆ and the vector effective length le = le le of the antenna. We determine the associated effective aperture as follows. Consider a co-polarized

10.13. EFFECTIVE APERTURE

197

e i incident on the sinusoidally-varying plane wave E co antenna. Further, let e i = Eie ˆ E co

(10.136)

e i . The ˆ is the reference direction of E where e co co-polarized power density incident on the antenna is: i 2 E i Sco = (10.137) 2η where η is the wave impedance of the medium (e.g., ∼ = 377 Ω for an antenna in free space). In response, the time-average power delivered to the load is o 1 n (10.138) PR = Re VeL IeL∗ 2

where VeL and IeL are the potential and current phasors, respectively, at the load. We may determine VeL and IeL from the equivalent circuit model of Figure 10.20 using basic circuit theory. First, note that the magnitude and phase of the voltage source is:   e i · l e = E i le e ˆ · ˆle E (10.139) co

Similarly, IeL is the voltage source output divided by the total series resistance: IeL =

PR =

1 i 2 RL E le 2 2 |ZA + ZL |

VeL = E le i



ZL ZA + ZL

(10.141)

(10.143)

Let us identify the real and imaginary components of ZA and ZL explicitly, as follows: ZA = RA + jXA

(10.144)

ZL = RL + jXL

(10.145)

Further, let us assume that loss internal to the antenna is negligible, so RA ≈ Rrad . If ZL is conjugate matched for maximum power transfer, then ∗ ZL = ZA = Rrad − jXA . Therefore: 1 i 2 Rrad E le ∗ |2 2 |ZA + ZA 2 Rrad 1 = E i le 2 2 |2Rrad | i 2 E le = (10.146) 8Rrad

PR,max =

(10.140)

Next, note that the load impedance creates a voltage divider with the source (antenna) impedance such that

(10.142)

Substituting these expressions into Equation 10.138, we obtain:

ˆ, and since we have defined ˆle to be equal to e e i · l e = E i le E co

E i le ZA + ZL

Now using Equations 10.135, 10.137, and 10.146, we find: i 2 E le /8Rrad (10.147) Ae = 2 |E i | /2η which reduces to:

2

Ae =

c S. Lally CC BY-SA 4.0 (modified)

η |le | 4Rrad

(10.148)

It is not surprising that effective aperture should be proportional to the square of the magnitude of effective length. However, we see that the medium and the radiation resistance of the antenna also play a role.

Figure 10.20: Equivalent circuit model for an antenna Effective length and radiation resistance are easy to e i , termi- calculate for wire antennas, so it is natural to use in the presence of an incident electric field E nated into a load impedance ZL . Equation 10.148 to calculate the effective apertures of

198

CHAPTER 10. ANTENNAS

wire antennas. For the electrically-short dipole (ESD) of length L, le ≈ (L/2) sin θ and 2 Rrad ≈ 20π 2 (L/λ) . Thus, we find the effective aperture assuming free space (i.e., η = η0 ) is:

Consider the scenario depicted in Figure 10.21. In this scenario, the antenna is completely enclosed in a chamber whose walls do not affect the behavior of the antenna and which have uniform temperature T . The load (still conjugate matched to the antenna) is 2 Ae ≈ 0.119λ2 |sin θ| (lossless ESD) (10.149) completely enclosed by a separate but identical chamber, also at temperature T . Remarkably, the effective aperture of the ESD does not depend on its length. In fact, it depends only on Consider the load chamber first. Heat causes random the frequency of operation. acceleration of constituent charge carriers in the load, giving rise to a randomly-varying current at the load Also worth noting is that maximum effective aperture terminals. (This is known as Johnson-Nyquist noise.) (i.e., the effective aperture in the θ = π/2 direction) is This current, flowing through the load, gives rise to a potential; therefore, the load – despite being a passive Ae ≈ 0.119λ2 (lossless ESD, max.) (10.150) device – is actually a source of power. The power associated with this noise is: Dipoles which are not electrically-short exhibit radiation resistance which is proportional to Lp , where p > 2. Therefore, the effective aperture of non-electrically-short dipoles does increase with L, as expected. However, this increase is not dramatic unless L is significantly greater than λ/2. Here’s an example: Example 10.10. Effective aperture of a half-wave dipole. The electrically-thin half-wave dipole exhibits radiation resistance ∼ = 73 Ω and effective length λ/π. Assuming the dipole is lossless and in free space, Equation 10.148 yields: Ae ≈ 0.131λ2 (half-wave dipole, max.) (10.151) Again, this is the effective aperture for a wave incident from broadside to the dipole.

Pload = kT B

(10.152)

where k ∼ = 1.38 × 10−23 J/K is Boltzmann’s constant and B is the bandwidth within which Pload is measured. Similarly, the antenna is a source of noise power Pant . Pant can also be interpreted as captured thermal radiation – that is, electromagnetic waves stimulated by the random acceleration of charged particles comprising the chamber walls. These waves radiate from the walls and travel to the antenna. The Rayleigh-Jeans law of thermodynamics tells us that this radiation should have power density (i.e., SI base units of W/m2 ) equal to 2kT B λ2

(10.153)

Effective aperture via thermodynamics. A useful insight into the concept of effective aperture can be deduced by asking the following question: How much power is received by an antenna when waves of equal power density arrive from all directions simultaneously? This question is surprisingly easy to answer using basic principles of thermodynamics. Chetvorno, public domain (modified) Thermodynamics is a field of physics that addresses heat and its relation to radiation and electrical energy. Figure 10.21: Thermodynamic analysis of the power No previous experience with thermodynamics is exchanged between an antenna and its load. assumed in this derivation.

10.13. EFFECTIVE APERTURE

199

per steradian of solid angle.5 The total power accessible to the antenna is one-half this amount, since an antenna is sensitive to only one polarization at a time, whereas the thermal radiation is equally distributed among any two orthogonal polarizations. PA is the remaining power, obtained by integrating over all 4π steradians. Therefore, the antenna captures total power equal to6   I 1 2kT ′ ′ Pant = Ae (θ , φ ) B sin θ′ dθ′ dφ′ 2 λ2 I  kT B Ae (θ′ , φ′ ) sin θ′ dθ′ dφ′ = λ2 (10.154) Given the conditions of the experiment, and in particular since the antenna chamber and the load chamber are at the same temperature, the antenna and load are in thermodynamic equilibrium. A consequence of this equilibrium is that power Pant captured by the antenna and delivered to the load must be equal to power Pload generated by the load and delivered to the antenna; i.e., Pant = Pload

(10.155)

Combining Equations 10.152, 10.154, and 10.155, we obtain:  I kT B Ae (θ′ , φ′ ) sin θ′ dθ′ dφ′ = kT B λ2 (10.156) which reduces to: I Ae (θ′ , φ′ ) sin θ′ dθ′ dφ′ = λ2 (10.157) Let hAe i be the mean effective aperture of the antenna; i.e., Ae averaged over all possible directions. This average is simply Equation 10.157 divided by the 4π sr of solid angle comprising all possible directions. Thus: hAe i =

λ2 4π

(10.158)

Recall that a isotropic antenna is one which has the same effective aperture in every direction. Such 5 Full disclosure: This is an approximation to the exact expression, but is very accurate at radio frequencies. See “Additional Reading” at the end of this section for additional details. 6 Recall sin θ ′ dθ ′ dφ′ is the differential element of solid angle.

antennas do not exist in practice, but the concept of an isotropic antenna is quite useful as we shall soon see. Since the effective aperture of an isotropic antenna must be the same as its mean effective aperture, we find: λ2 ≈ 0.080λ2 (isotropic antenna) 4π (10.159) Note further that this must be the minimum possible value of the maximum effective aperture for any antenna. Ae =

Effective aperture via reciprocity. The fact that all antennas have maximum effective aperture greater than that of an isotropic antenna makes the isotropic antenna a logical benchmark against which to compare the effective aperture of antennas. We can characterize the effective aperture of any antenna as being some unitless real-valued constant D times the effective aperture of an isotropic antenna; i.e.: Ae , D

λ2 (any antenna) 4π

(10.160)

where D must be greater than or equal to 1. Astute readers might notice that D seems a lot like directivity. Directivity is defined in Section 10.7 as the factor by which a transmitting antenna increases the power density of its radiation over that of an isotropic antenna. Let us once again consider the ESD, for which we previously determined (via the effective length concept) that Ae ∼ = 0.119λ2 in the direction in which it is maximum. Applying Equation 10.160 to the ESD, we find: D , Ae

4π ∼ = 1.50 (ESD, max.) λ2

(10.161)

Sure enough, D is equal to the directivity of the ESD in the transmit case. This result turns out to be generally true; that is: The effective aperture of any antenna is given by Equation 10.160 where D is the directivity of the antenna when transmitting. A derivation of this result for the general case is possible using analysis similar to the thermodynamics

200 presented earlier, or using the reciprocity theorem developed in Section 10.10. The fact that effective aperture is easily calculated from transmit directivity is an enormously useful tool in antenna engineering. Without this tool, determination of effective aperture is limited to direct measurement or calculation via effective length; and effective length is generally difficult to calculate for antennas which are not well-described as distributions of current along a line. It is usually far easier to calculate the directivity of an antenna when transmitting, and then use Equation 10.160 to obtain effective aperture for receive operation. Note that this principle is itself an expression of reciprocity. That is, one may fairly say that “directivity” is not exclusively a transmit concept nor a receive concept, but rather is a single quantifiable characteristic of an antenna that applies in both the transmit and receive case. Recall that radiation patterns are used to quantify the way (transmit) directivity varies with direction. Suddenly, we have found that precisely the same patterns apply to the receive case! Summarizing: As long as the conditions required for formal reciprocity are satisfied, the directivity of a receiving antenna (defined via Equation 10.160) is equal to directivity of the same antenna when transmitting, and patterns describing receive directivity are equal to those for transmit directivity.

Finally, we note that the equivalence of transmit and receive directivity can, again via Equation 10.160, be used to define an effective aperture for the transmit case. In other words, we can define the effective aperture of a transmitting antenna to be λ2 /4π times the directivity of the antenna. There is no new physics at work here; we are simply taking advantage of the fact that the concepts of effective aperture and directivity describe essentially the same characteristic of an antenna, and that this characteristic is the same for both transmit and receive operation. Additional Reading: • “Antenna aperture” on Wikipedia.

CHAPTER 10. ANTENNAS • “Boltzmann constant” on Wikipedia. • “Rayleigh-Jeans law” on Wikipedia. • “Johnson-Nyquist noise” on Wikipedia. • “Thermodynamics” on Wikipedia.

10.14. FRIIS TRANSMISSION EQUATION

10.14 Friis Transmission Equation [m0219]

A common task in radio systems applications is to determine the power delivered to a receiver due to a distant transmitter. The scenario is shown in Figure 10.22: A transmitter delivers power PT to an antenna which has gain GT in the direction of the receiver. The receiver’s antenna has gain GR . As always, antenna gain is equal to directivity times radiation efficiency, so GT and GR account for losses internal to the antenna, but not losses due to impedance mismatch.

201 aperture Ae of the receive antenna: i PR,max = Ae Sco PT = Ae GT 4πR2

(10.164)

This assumes that the receive antenna is co-polarized with the incident electric field, and that the receiver is conjugate-matched to the antenna. The effective aperture can also be expressed in terms of the gain GR of the receive antenna: Ae =

λ2 GR 4π

(10.165)

Thus, Equation 10.164 may be written in the following form:

 2 A simple expression for PR can be derived as follows. λ PR,max = PT GT GR (10.166) First, let us assume “free space conditions”; that is, let 4πR us assume that the intervening terrain exhibits negligible absorption, reflection, or other scattering of This is the Friis transmission equation. Summarizing: the transmitted signal. In this case, the spatial power density at range R from the transmitter which radiates this power through a lossless and isotropic antenna The Friis transmission equation (Equawould be: tion 10.166) gives the power delivered to a PT conjugate-matched receiver in response to a dis(10.162) 4πR2 tant transmitter, assuming co-polarized antennas and free space conditions. that is, total transmitted power divided by the area of a sphere of radius R through which all the power must flow. The actual power density S i is this amount 2 The factor (λ/4πR) appearing in the Friis times the gain of the transmit antenna, i.e.: transmission equation is referred to as free space path gain. More often this is expressed as the reciprocal PT i S = GT (10.163) quantity: 4πR2  −2 λ The maximum received power is the incident Lp , (10.167) 4πR co-polarized power density times the effective which is known as free space path loss. Thus, Equation 10.166 may be expressed as follows: PR,max = PT GT L−1 p GR

(10.168)

The utility of the concept of path loss is that it may also be determined for conditions which are different from free space. The Friis transmission equation still applies; one simply uses the appropriate (and probably significantly different) value of Lp . Figure 10.22: Radio link accounting for antenna gains A common misconception is that path loss is equal to and spreading of the transmitted wave. the reduction in power density due to spreading along

CHAPTER 10. ANTENNAS

202 the path between antennas, and therefore this “spreading loss” increases with frequency. In fact, the reduction in power density due to spreading between any two distances R1 < R2 is:  2 PT /4πR12 R1 = (10.169) PT /4πR22 R2

that the height and high directivity of the antennas yield conditions sufficiently close to free space. We further assume conjugate-matching at the receiver, and that the antennas are co-polarized. Under these conditions, PR = PR,max and Equation 10.166 applies. We find:

which is clearly independent of frequency. The path PR,max PT ≥ loss Lp , in contrast, depends only on the total distance 2 GT (λ/4πR) GR R and does depend on frequency. The dependence on ∼ = 2.26 × 10−7 W frequency reflects the dependence of the effective ∼ aperture on wavelength. Thus, path loss is not loss in = 2.26 × 10−4 mW the traditional sense, but rather accounts for a ∼ = −36.5 dBm combination of spreading and the λ2 dependence of effective aperture that is common to all receiving antennas. Additional Reading: Finally, note that Equation 10.168 is merely the simplest form of the Friis transmission equation. Commonly encountered alternative forms include forms in which GT and/or GR are instead represented by the associated effective apertures, and forms in which the effects of antenna impedance mismatch and/or cross-polarization are taken into account. Example 10.11. 6 GHz point-to-point link. Terrestrial telecommunications systems commonly aggregate large numbers of individual communications links into a single high-bandwidth link. This is often implemented as a radio link between dish-type antennas having gain of about 27 dBi (that’s dB relative to a lossless isotropic antenna) mounted on very tall towers and operating at frequencies around 6 GHz. Assuming the minimum acceptable receive power is −120 dBm (that’s −120 dB relative to 1 mW; i.e., 10−15 W) and the required range is 30 km, what is the minimum acceptable transmit power? Solution. From the problem statement: GT = GR = 1027/10 ∼ = 501 λ=

c ∼ 3 × 108 m/s ∼ = = 5.00 cm f 6 × 109 Hz

(10.170) (10.171)

R = 30 km, and PR ≥ 10−15 W. We assume

(10.172)

• “Free-space path loss” on Wikipedia. • “Friis transmission equation” on Wikipedia. [m0211]

10.14. FRIIS TRANSMISSION EQUATION

203

Image Credits c Offaperry (S. Lally), https://commons.wikimedia.org/wiki/File:Standing Wave Creation.svg, Fig. 10.1: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), https://commons.wikimedia.org/wiki/File:Electrically-Short Dipole.svg, Fig. 10.2: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), Fig. 10.3: https://commons.wikimedia.org/wiki/File:Zero-Length Dipole Open Circuit.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), Fig. 10.4: https://commons.wikimedia.org/wiki/File:Transmitting Antenna Circuit.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), Fig. 10.5: https://commons.wikimedia.org/wiki/File:Magnitude of the Radiated Field.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), https://commons.wikimedia.org/wiki/File:FHplaneMag.svg, Fig. 10.6: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c T. Truckle, https://en.wikipedia.org/wiki/File:Sidelobes en.svg, Fig. 10.7: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). Modified. c Offaperry (S. Lally), https://commons.wikimedia.org/wiki/File:Half-Power Beamwidth.svg, Fig. 10.8: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), https://commons.wikimedia.org/wiki/File:Antenna Equivalent Circuit.svg, Fig. 10.9: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). Modified. Fig. 10.10: Inductiveload, https://commons.wikimedia.org/wiki/File:Two Port Circuit.svg, public domain. Modified. c Sevenchw (C. Wang), Fig. 10.11: https://commons.wikimedia.org/wiki/File:An electromagnetic system consisting of a current distribution radiating an electric field.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 10.12: https://commons.wikimedia.org/wiki/File:An electromagnetic system consisting of a different current distribution radiating an electric field.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 10.13: https://commons.wikimedia.org/wiki/File:A two-port consisting of two dipole antennas.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 10.14: https://commons.wikimedia.org/wiki/File:The port 1 antenna a transmitting and b receiving.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/).

204

CHAPTER 10. ANTENNAS

c Sevenchw (C. Wang), Fig. 10.15: https://commons.wikimedia.org/wiki/File:Thin straight dipole respond to incident plane wave.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 10.16: https://commons.wikimedia.org/wiki/File:Dipole of interest driven by current.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Sevenchw (C. Wang), Fig. 10.17: https://commons.wikimedia.org/wiki/File:Hertzian dipole drive by current.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). c Offaperry (S. Lally), Fig. 10.18: https://commons.wikimedia.org/wiki/File:Antenna Equivalent Circuit.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). Fig. 10.19: Inductiveload, https://commons.wikimedia.org/wiki/File:Two Port Circuit.svg, public domain. Modified. c Offaperry (S. Lally), Fig. 10.20: https://commons.wikimedia.org/wiki/File:Antenna Equivalent Circuit.svg, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). Modified. Fig. 10.21: Chetvorno, https://en.wikipedia.org/wiki/File:Antenna and resistor in cavity.svg, public domain. Modified.

Appendix A

Constitutive Parameters of Some Common Materials A.1 Permittivity of Some Common Materials [m0135]

The values below are relative permittivity ǫr , ǫ/ǫ0 for a few materials that are commonly encountered in electrical engineering applications, and for which permittivity emerges as a consideration. Note that “relative permittivity” is sometimes referred to as dielectric constant. Here we consider only the physical (real-valued) permittivity, which is the real part of the complex permittivity (typically indicated as ǫ′ or ǫ′r ) for materials exhibiting significant loss. Permittivity varies significantly as a function of frequency. The values below are representative of frequencies from a few kHz to about 1 GHz. The values given are also representative of optical frequencies for materials such as silica that are used in optical applications. Permittivity also varies as a function of temperature. In applications where precision better than about 10% is required, primary references accounting for frequency and temperature should be consulted. The values presented here are gathered from a variety of references, including those indicated in “Additional References.” Free Space (vacuum): ǫr , 1

Solid Dielectrics: Material Styrofoam1 Teflon2 Polyethylene Polypropylene Silica Polystyrene Polycarbonate Rogers RO3003 FR4 (glass epoxy laminate)

ǫr 1.1 2.1 2.3 2.3 2.4 2.6 2.8 3.0 4.5

Common uses

coaxial cable optical fiber3

PCB substrate PCB substrate

1

Properly known as extruded polystyrene foam (XPS). 2 Properly known as polytetrafluoroethylene (PTFE). 3 Typically doped with small amounts of other materials to slightly raise or lower the index of √ refraction (= ǫr ). Non-conducting spacing materials used in discrete capacitors exhibit ǫr ranging from about 5 to 50. Semiconductors commonly appearing in electronics – including carbon, silicon, geranium, indium phosphide, and so on – typically exhibit ǫr in the range 5–15. Glass exhibits ǫr in the range 4–10, depending on composition. Gasses, including air, typically exhibit ǫr ∼ = 1 to within a tiny fraction of a percent. Liquid water typically exhibits ǫr in the range 72–81. Distilled water exhibits ǫr ≈ 81 at room temperature, whereas sea water tends to be at the

c 2020 S.W. Ellingson CC BY SA 4.0. https://doi.org/10.21061/electromagnetics-vol-2 Electromagnetics Vol. 2.

206

APPENDIX A. CONSTITUTIVE PARAMETERS OF SOME COMMON MATERIALS

lower end of the range. Other liquids typically exhibit ǫr in the range 10–90, with considerable variation as a function of temperature and frequency. Animal flesh and blood consists primarily of liquid matter and so also exhibits permittivity in this range. Soil typically exhibits ǫr in the range 2.5–3.5 when dry and higher when wet. The permittivity of soil varies considerably depending on composition. Additional Reading: • CRC Handbook of Chemistry and Physics. • “Relative permittivity” on Wikipedia. • “Miscellaneous Dielectric Constants” on microwaves101.com.

A.2

Permeability of Some Common Materials

[m0136]

The values below are relative permeability µr , µ/µ0 for a few materials that are commonly encountered in electrical engineering applications, and for which µr is significantly different from 1. These materials are predominantly ferromagnetic metals and (in the case of ferrites) materials containing significant ferromagnetic metal content. Nearly all other materials exhibit µr that is not significantly different from that of free space. The values presented here are gathered from a variety of references, including those indicated in “Additional References” at the end of this section. Be aware that permeability may vary significantly with frequency; values given here are applicable to the frequency ranges for applications in which these materials are typically used. Also be aware that materials exhibiting high permeability are also typically non-linear; that is, permeability depends on the magnitude of the magnetic field. Again, values reported here are those applicable to applications in which these materials are typically used. Free Space (vacuum): µr , 1. Iron (also referred to by the chemical notation “Fe”) appears as a principal ingredient in many materials and alloys employed in electrical structures and devices. Iron exhibits µr that is very high, but which decreases with decreasing purity. 99.95% pure iron exhibits µr ∼ 200, 000. This decreases to ∼ 5000 at 99.8% purity and is typically below 100 for purity less than 99%. Steel is an iron alloy that comes in many forms, with a correspondingly broad range of permeabilities. Electrical steel, commonly used in electrical machinery and transformers when high permeability is desired, exhibits µr ∼ 4000. Stainless steel, encompassing a broad range of alloys used in mechanical applications, exhibits µr in the range 750–1800. Carbon steel, including a broad class of alloys commonly used in structural applications, exhibits µr on the order of 100.

A.3. CONDUCTIVITY OF SOME COMMON MATERIALS Ferrites include a broad range of ceramic materials that are combined with iron and various combinations of other metals and are used as magnets and magnetic devices in various electrical systems. Common ferrites exhibit µr in the range 16–640.

207

A.3 Conductivity of Some Common Materials

[m0137]

Additional Reading: • CRC Handbook of Chemistry and Physics. • “Magnetic Materials” on microwaves101.com. • “Permeability (electromagnetism)” on Wikipedia. • “Iron” on Wikipedia. • “Electrical steel” on Wikipedia. • “Ferrite (magnet)” on Wikipedia.

The values below are conductivity σ for a few materials that are commonly encountered in electrical engineering applications, and for which conductivity emerges as a consideration. Note that materials in some applications are described instead in terms of resistivity, which is simply the reciprocal of conductivity. Conductivity may vary significantly as a function of frequency. The values below are representative of frequencies from a few kHz to a few GHz. Conductivity also varies as a function of temperature. In applications where precise values are required, primary references accounting for frequency and temperature should be consulted. The values presented here are gathered from a variety of references, including those indicated in “Additional References” at the end of this section. Free Space (vacuum): σ , 0. Commonly encountered elements: Material σ (S/m) Copper 5.8 × 107 Gold 4.4 × 107 Aluminum 3.7 × 107 Iron 1.0 × 107 Platinum 0.9 × 107 Carbon 1.3 × 105 Silicon 4.4 × 10−4 Water exhibits σ ranging from about 6 µS/m for highly distilled water (thus, a very poor conductor) to about 5 S/m for seawater (thus, a relatively good conductor), varying also with temperature and pressure. Tap water is typically in the range of 5–50 mS/m, depending on the level of impurities present. Soil typically exhibits σ in the range 10−4 S/m for dry soil to about 10−1 S/m for wet soil, varying also due to chemical composition.

208

APPENDIX A. CONSTITUTIVE PARAMETERS OF SOME COMMON MATERIALS

Non-conductors. Most other materials that are not well-described as conductors or semiconductors and are dry exhibit σ < 10−12 S/m. Most materials that are considered to be insulators, including air and common dielectrics, exhibit σ < 10−15 S/m, often by several orders of magnitude. Additional Reading: • CRC Handbook of Chemistry and Physics. • “Conductivity (electrolytic)” on Wikipedia. • “Electrical resistivity and conductivity” on Wikipedia. • “Soil resistivity” on Wikipedia.

Appendix B

Mathematical Formulas B.1 Trigonometry

B.2 Vector Operators

[m0138]

[m0139]

ejθ = cos θ + j sin θ

This section contains a summary of vector operators expressed in each of the three major coordinate systems:

 1 jθ e + e−jθ 2

cos θ =

sin θ =

 1 jθ e − e−jθ j2

(B.1)

(B.2)

• cylindrical (ρ,φ,z) (B.3)

1 1 cos θ = + cos 2θ 2 2

(B.4)

1 1 − cos 2θ 2 2

(B.5)

2

sin2 θ =

(B.6)

cos (a ± b) = cos a cos b ∓ sin a sin b

(B.7)

cosh θ =

 1 +θ e + e−θ 2

Associated basis vectors are identified using a caret (ˆ) over the symbol. The vector operand A is expressed in terms of components in the basis directions as follows:

ˆ φ+z ˆ Az • cylindrical: A = ρˆAρ + φA ˆ θ + φA ˆ φ • spherical: A = ˆrAr + θA Gradient Gradient in Cartesian coordinates:

Hyperbolic trigonometric functions:  1 +θ e − e−θ 2

• spherical (r,θ,φ)

ˆ Ax + y ˆ Ay + z ˆ Az • Cartesian: A = x

sin (a ± b) = sin a cos b ± cos a sin b

sinh θ =

• Cartesian (x,y,z)

(B.8) ˆ ∇f = x

∂f ∂f ∂f ˆ ˆ +y +z ∂x ∂y ∂z

(B.10)

(B.9) Gradient in cylindrical coordinates: ∇f = ρˆ

∂f 1 ∂f ∂f ˆ + φˆ +z ∂ρ ρ ∂φ ∂z

c 2020 S.W. Ellingson CC BY SA 4.0. https://doi.org/10.21061/electromagnetics-vol-2 Electromagnetics Vol. 2.

(B.11)

APPENDIX B. MATHEMATICAL FORMULAS

210

Laplacian Laplacian in Cartesian coordinates:

Gradient in spherical coordinates: ∇f = ˆr

∂f 1 ∂f 1 ∂f + θˆ + φˆ ∂r r ∂θ r sin θ ∂φ

Divergence Divergence in Cartesian coordinates: ∇·A=

∂Ax ∂Ay ∂Az + + ∂x ∂y ∂z

Divergence in cylindrical coordinates: ∇·A=

1 ∂ 1 ∂Aφ ∂Az (ρAρ ) + + ρ ∂ρ ρ ∂φ ∂z

Divergence in spherical coordinates:  1 ∂ r2 Ar 2 r ∂r ∂ 1 (Aθ sin θ) + r sin θ ∂θ 1 ∂Aφ + r sin θ ∂φ

∇·A=

Curl Curl in Cartesian coordinates:   ∂Az ∂Ay ˆ ∇×A= x − ∂y ∂z   ∂Ax ∂Az ˆ +y − ∂z ∂x   ∂Ay ∂Ax ˆ +z − ∂x ∂y Curl in cylindrical coordinates:   ∂Aφ 1 ∂Az − ∇ × A = ρˆ ρ ∂φ ∂z   ∂A ∂A ρ z + φˆ − ∂z ∂ρ   1 ∂ ∂Aρ ˆ +z (ρAφ ) − ρ ∂ρ ∂φ

(B.12) ∇2 f =

∂2f ∂2f ∂2f + + ∂x2 ∂y 2 ∂z 2

Laplacian in cylindrical coordinates:   1 ∂ ∂f 1 ∂2f ∂2f (B.13) ∇2 f = ρ + 2 2+ 2 ρ ∂ρ ∂ρ ρ ∂φ ∂z Laplacian in spherical coordinates:   (B.14) 1 ∂ ∂f ∇2 f = 2 r2 r ∂r ∂r   1 ∂ ∂f + 2 sin θ r sin θ ∂θ ∂θ 1 ∂2f + 2 2 r sin θ ∂φ2 (B.15)

(B.16)

(B.17)

Curl in spherical coordinates:   1 ∂ ∂Aθ ∇ × A = ˆr (Aφ sin θ) − r sin θ ∂θ ∂φ   1 1 ∂Ar ∂ + θˆ − (rAφ ) r sin θ ∂φ ∂r   1 ∂ ∂Ar + φˆ (rAθ ) − (B.18) r ∂r ∂θ

(B.19)

(B.20)

(B.21)

B.3. VECTOR IDENTITIES

211

B.3 Vector Identities [m0140]

Algebraic Identities A · (B × C) = B · (C × A) = C · (A × B) A × (B × C) = B (A · C) − C (A · B)

(B.22) (B.23)

Identities Involving Differential Operators ∇ · (∇ × A) = 0

(B.24)

∇ × (∇f ) = 0 ∇ × (f A) = f (∇ × A) + (∇f ) × A

∇ · (A × B) = B · (∇ × A) − A · (∇ × B) 2

∇ · (∇f ) = ∇ f

(B.25) (B.26) (B.27) (B.28)

2

∇ × ∇ × A = ∇ (∇ · A) − ∇ A 2

∇ A = ∇ (∇ · A) − ∇ × (∇ × A)

(B.29) (B.30)

Divergence Theorem: Given a closed surface S enclosing a contiguous volume V, Z I (∇ · A) dv = A · ds (B.31) V

S

where the surface normal ds is pointing out of the volume. Stokes’ Theorem: Given a closed curve C bounding a contiguous surface S, Z I (∇ × A) · ds = A · dl (B.32) S

C

where the direction of the surface normal ds is related to the direction of integration along C by the “right hand rule.”

Appendix C

Physical Constants [m0141]

The speed of light in free space (c), which is the phase velocity of any electromagnetic radiation in free space, is ∼ = 2.9979 × 108 m/s. This is commonly rounded up to 3 × 108 m/s. This rounding incurs error of ∼ = 0.07%, which is usually much less than other errors present in electrical engineering calculations. The charge of an electron is ∼ = −1.602 × 10−19 C. The constant e , +1.602176634 × 10−19 C is known as the “elementary charge,” so the charge of the electron is said to be −e. The permittivity of free space (ǫ0 ) is ∼ = 8.854 × 10−12 F/m. The permeability of free space (µ0 ) is 4π × 10−7 H/m. The wave impedance of free space (η0 ) is the ratio of the magnitude of the electric field intensity to that of p the magnetic field intensity in free space and is µ0 /ǫ0 ∼ = 376.7 Ω. This is also sometimes referred to as the intrinsic impedance of free space. Boltzmann’s constant is ∼ = 1.381 × 10−23 J/K, the amount of energy associated with a change of one degree of temperature. This is typically assigned the symbol k (unfortunately, the same symbol often used to represent wavenumber).

c 2020 S.W. Ellingson CC BY SA 4.0. https://doi.org/10.21061/electromagnetics-vol-2 Electromagnetics Vol. 2.

Index acceptance angle, 140 aluminum, 43, 46, 207 Ampere’s law general form, 7, 8, 26, 34, 105, 146, 154, 155, 159 magnetostatics, 6, 18, 124 antenna electrically-short dipole (ESD), 155–157, 159, 161, 168–172, 175–177, 198 folded half-wave dipole, 136 half-wave dipole, 136, 159, 161–162, 198 isotropic, 199, 201 microstrip patch, 174 quarter-wave monopole, 136 yagi, 181 antenna aperture, 196 antenna factor, 183 attenuation, 129–133, 135 attenuation rate, 40–41, 44 balun, 136 binomial series, 42 Biot-Savart law, 18–20, 154 Boltzmann’s constant, 198, 212 boundary conditions, 5, 6, 8, 187, 190 Brewster’s angle, 86 carbon, 207 CGS (system of units), 2 characteristic impedance, 122, 124 charge density line, 5 surface, 5 volume, 5 cladding, 138 conductivity of common materials, 207–208 conductor good, 43–45, 48, 169 perfect, 6, 43, 48, 50, 58 poor, 41–43

cone of acceptance, 140 constitutive relationships, 7 copper, 43, 207 Coulomb force, 21 Coulomb’s law, 18 coupling, 196 critical angle, 88 curl, 210 current, 5, 48 current density surface, 5, 163 volume, 5, 163 current moment, 20, 150, 152, 162 cutoff frequency, 100, 106, 117, 123 cyclotron motion, 11 dB, see decibel dBm, 40 decibel, 39–40 delay spread, 143 dielectric (material), 123 examples, 205 dielectric constant, 205 Dirac delta function, 145, 150, 162 directivity, 177–179, 199 dispersion, 54, 95, 102, 118, 141 displacement current, 7, 34 divergence, 210 divergence theorem, 26, 211 E-plane, 180 effective aperture, 196–200 effective height, 183 effective length, 183, 190 electric field intensity, 5, 11 electric flux density, 5 electrical length, 154, 168 electrically-short dipole, see antenna electrically-short dipole (ESD), 178 electron, 212 electrostatics (description), 5

213

INDEX

214 English system of units, 2 evanescent waves, see waves far field, 153, 161, 167 Faraday’s law, 7, 23 ferrite, 207 fiber optics, 88 flux electric, 7 magnetic, 6, 7 force, 20 FR4, 41, 127, 128, 136, 205 Friis transmission equation, 201–202 gain power, 39 voltage, 40 Gauss’ law electric field, 5, 18 magnetic field, 6, 148 geranium, 205 glass, 205 gold, 43, 207 good conductor, 50 Goos-H¨anchen effect, 89 gradient, 209 group velocity, 95–97, 102, 118 H-plane, 180 half-power beamwidth (HPBW), 182 half-wave matching (slab), 65–66 Hertzian dipole, 147, 152–156, 159, 166, 169, 176, 191 homogeneity, 8 homogeneous (media), 8 impedance surface, 54 wire, 50–54 independence of path, 5 index of refraction, 59, 81, 138, 205 indium phosphide, 205 inductance, 53 equivalent, 53 inverse square law, 18 iron, 206, 207 isotropic, 182 isotropic (media), 9 isotropy, 9 Jefimenko’s equations, 18

Johnson-Nyquist noise, 198 joule heating, 27 Joule’s law, 27 Kirchoff’s voltage law electrostatics, 5 Laplace’s Equation, 133 Laplacian (operator), 210 lever arm, 15 linear (media), 9 linearity, 9 Lorentz force, 11 Lorentz reciprocity theorem, 187–188 Lorenz gauge condition, 149 loss resistance, 174 loss tangent, 30, 34–35, 41, 43 magnetic field intensity, 6 magnetic flux density, 6, 11, 148 magnetic vector potential, 147–152, 159 magnetostatics (description), 6 materials (properties), 8 materials, magnetic, 6 Maxwell’s equations, 7, 145, 147 differential phasor form, 7, 30, 108, 187 source-free lossless region, 8 static differential form, 6 static integral form, 6 Maxwell-Faraday equation, 7, 8, 26, 102 metric system, 2 microstrip, 108, 123–129, 136 mode, 100, 106, 112, 116 motional emf, 23 motor, 13, 17 mutual coupling, 196 notation, 3 numerical aperture, 140 Ohm’s law, 6, 25, 27, 30, 50 ohmic loss, 27, 34 omnidirectional, 182 optical fiber, 138–143 parallel-plate waveguide, see waveguide, see waveguide, see waveguide, see waveguide, see waveguide path gain, 201 path loss, 201 pattern (radiation), 179–182, 200

INDEX normalized, 180 permeability, 6 of common materials, 206–207 relative, 6, 206 permittivity, 5 complex-valued, 30, 33–34 effective, 128 of common materials, 205–206 relative, 5, 205 phase velocity, 95–97, 118 in microstrip, 128 phasor, 7 plane of incidence, 70 plane wave relationships, 36, 73, 77, 107, 154, 157, 161 plasma, 147 platinum, 207 polarization, 179–182 polarizing angle, 86 polycarbonate, 205 polyethylene, 42, 205 polypropylene, 205 polystyrene, 205 polytetrafluoroethylene, 205 potential difference, 21 power handling, see transmission line Poynting vector, 27–29, 37 Poynting’s theorem, 25–28, 146 printed circuit board (PCB), 123, 128, 174 material, 205 prism, 82 propagation constant, 150 attenuation, 32, 36, 40, 46, 150 phase, 30, 31, 36, 150 quarter-wave matching (slab), 66–67 radiation condition, 151 radiation efficiency, 169, 174, 178 radiation resistance, 174, 175 radome, 60, 64 ray-fixed coordinates, 68, 70 Rayleigh-Jeans law, 198 reciprocity, 186–190 rectangular waveguide, see waveguide reference phase, 67 reference polarization, 67 reflection coefficient, 57 refraction, 81 resistivity, 207

215 RG-59, 52, 53, 131 right hand rule magnetostatics, 19, 124 Stokes’ theorem, 211 sampling function, 145, 150 semiconductor (material), 205 separation of variables, 111, 114 SI (system of units), 2 silica, 205 silicon, 205, 207 skin depth, 45–46, 50, 51, 129 skin effect, 49, 53 slab, 60–67 Snell’s law, 81, 83, 85, 88, 92, 140 soil, 206, 207 speed of light, 212 standing wave, 63, 161 steel, 206 Stokes’ theorem, 211 stripline, 123 styrofoam, 205 superposition, 9, 108, 111, 114, 146, 163, 166 surface wave, see waves Teflon, 205 thermal radiation, 198 thermodynamic equilibrium, 199 thermodynamics, 198 time-harmonic, 7, 147 time-invariance (media), 9 torque, 15 total internal reflection, 82, 84, 86, 88–90, 138 transducer, 166 transmission line analogy for wave propagation, 58, 63 coaxial, 52, 53, 103, 121, 129–135 differential, 121 lossy, 36 lumped-element model, 122, 124 microstrip, see microstrip parallel wire, 121–123 power handling, 133–136 single-ended, 121, 123, 136 stripline, 123 transverse electromagnetic (TEM), 32, 124 transverse electric, 70–75, 83–84, 110, 113–116 transverse electromagnetic (TEM), 71, 103, 110, 122 transverse magnetic, 70–71, 76–80, 85–87, 110–113 twin lead, 121

216 units, 1–2 vector arithmetic, 211 identity, 211 position-free, 15 vector effective length, 183, 190, 192, 195, 196 water, 205, 207 wave equation electromagnetic, 36, 147, 149, 150 magnetic vector potential, 149, 150 source-free lossless region, 8 source-free lossy region, 30–32 wave impedance, 8, 37, 42, 154, 212 waveguide parallel plate, 97–108 rectangular, 110–119 wavelength in microstrip, 128 waves evanescent, 90–92 plane, 8, 36–37 power density, 8, 37 surface, 92 transmission line analogy, 36 unidirectional, 108–110 work, 20 yagi, see antenna

INDEX

Electromagnetics, volume 2, by Steven W. Ellingson is a 216-page peer-reviewed open textbook designed especially for electrical engineering students in the third year of a bachelor of science degree program. It is intended as the primary textbook for the second semester of a two-semester undergraduate engineering electromagnetics sequence. The book addresses magnetic force and the Biot-Savart law; general and lossy media; parallel plate and rectangular waveguides; parallel wire, microstrip, and coaxial transmission lines; AC current flow and skin depth; reflection and transmission at planar boundaries; fields in parallel plate, parallel wire, and microstrip transmission lines; optical fiber; and radiation and antennas. This textbook is part of the Open Electromagnetics Project led by Steven W. Ellingson at Virginia Tech (https://www.faculty.ece.vt.edu/swe/oem). The project goal is to create publicly available, no-cost, openly licensed content for courses in engineering electromagnetics. The project is motivated by two things: lowering the cost of learning materials for students and giving faculty the freedom to adopt, modify, and improve their educational resources. For each book in the series, the following are freely available for all to use: openly licensed and accessible PDFs, problem sets and solutions, slides of figures, editable LaTeX source files, and information for the book’s collaborator portal and listserv. Books in this series Electromagnetics, volume 1, https://doi.org/10.21061/electromagnetics-vol-1 Electromagnetics, volume 2, https://doi.org/10.21061/electromagnetics-vol-2

Advance Praise for Electromagnetics “I commend the author for the hard work and generosity required to create a book like this and to make it available for free. The overall format is pleasing and the material is generally rigorous and complete. Well done.” — Karl Warnick, Brigham Young University “I really liked some of the ways the author related the math to physics and explained why the math made sense. … The figures are generally quite good. … The examples in the book are simple and straightforward.” — Randy Haupt, Colorado School of Mines

This book is licensed with a Creative Commons Attribution-ShareAlike License 4.0.

Virginia Tech Publishing ∙ Blacksburg, Virginia

Cover Design: Robert Browder